US 10958840

I'll run several targeted searches, then synthesize. Note: I already have the authoritative patent text in-context, but I need to verify the claims and the litigation/docket identifiers.

Now let me pin down the Federal Circuit 2026 dockets and the PTAB proceedings.

US 10,958,840 B2 — verified summary

Bottom line: Everything below is anchored to the authoritative full text supplied in-context (Google Patents) plus live searches of USPTO PTAB/PTACTS filings, govinfo Federal Register, and the ITC's published notices. I did not have direct PACER/CAFC docket-system access, so the Federal Circuit entries are reported from secondary/aggregator sources and are flagged as such.

1. Bibliographic data

2. Abstract (verbatim)

"Visual content is captured by an image capture device during a capture duration. The image capture devices experiences change in position during the capture duration. The trajectory of the image capture device is smoothed based on a look ahead of the trajectory. A punchout of the visual content is determined based on the smoothed trajectory. The punchout of the visual content is used to generate stabilized visual content."

Note: the abstract and specification use "smoothed trajectory"; the closely related sibling patent US 10,341,564 / US 11,025,824 and the EPO family entry instead use the older "observed trajectory / capture trajectory" wording. (One EPO PISE record literally lists the abstract against "US10432864B1" — I am reporting that identifier as-is, without correcting it.)

3. Independent claims in plain language

The '840 patent has two independent claims: claim 1 (image capture system) and claim 16 (method). This is confirmed by Petitioner's expert declaration in the PTAB record: "The independent claims of the '840 patent are directed to an image capture device and a method that stabilizes videos. (EX1001, Claims 1, 16.)" Claims 1 and 16 include the limitation "determine a smoothed trajectory and one or more of a weight‑balance parameter, a low‑light high‑pass parameter, and/or a stickiness parameter."

Claim 1 — "image capture system" (reconstructed from the patent PDF; the final elements are verbatim, the preamble/joint elements are partly summarized):

A camera system with a housing carrying an optical element, image sensor, position sensor and processor(s), wherein:

1. Capture & sensing — the image sensor produces visual output defining visual content having a field of view during a capture duration; the position sensor produces position information characterizing rotational positions of the housing at different moments in the capture duration (the specification adds that position info may also characterize translational positions, and is derived independent of the imagery, e.g. gyro/IMU).
2. Trajectory — determine a trajectory of the housing from the position information, reflecting the housing's rotational positions at those moments, with a first portion (first moment) and a second, later portion (second moment).
3. Smoothed trajectory (the core) — determine a smoothed trajectory based on a look‑ahead of the trajectory AND one or more of a weight‑balance parameter, a low‑light high‑pass parameter, and/or a stickiness parameter. The look‑ahead uses a later portion of the trajectory to determine an earlier portion of the smoothed trajectory, so a part of the smoothed trajectory corresponding to the first portion is determined from the second portion. The smoothed trajectory has smoother changes in rotational position than the raw trajectory. The look‑ahead must use a temporal horizon of motion that enables approximation of intentional motion while causing a delay in determining the smoothed trajectory.
4. Viewing window — determine placement of a viewing window relative to the field of view based on the smoothed trajectory, the window defining extents of the visual content to be included in a punchout.
5. Stabilized output — generate stabilized visual content of a video based on the viewing window, the stabilized content including the punchout, "wherein inclusion in the stabilized visual content of the one or more extents of the visual content within the viewing window effectuates stabilization of the visual content via selective cropping."

Claim 16 — method mirrors claim 1 in method form: generating the visual output signal; generating the position output signal; determining the trajectory; determining the smoothed trajectory from the look‑ahead (plus the parameter(s) and temporal‑horizon limitation); "determining, by the one or more processors, placement of the viewing window for the visual content with respect to the field of view…"; and generating the stabilized visual content including the punchout.

Dependent‑claim map (claims seen verbatim in retrieved sources):

Claims 4, 7, 12, and 17–21 were not retrieved verbatim in this session — I will not guess at their content. (Claim 12 and 17–21 are method‑claim dependents in the 16–21 block.)

Specification highlights relevant to claim construction: objective function = L2‑norm of interframe angular velocity + angular acceleration, weighted by constants; iterative fine‑scale then coarse‑scale smoothing with a projection step enforcing the crop/margin constraint; parameter ranges 0–1 for weight‑balance, low‑light high‑pass, and stickiness; smoothing values S1–S4 as a function of exposure time (T1/T2 thresholds) and camera motion (FIG. 7); post‑capture stabilization may be performed by a remote processor.

4. Litigation and PTAB status (patent‑specific)

ITC — Inv. No. 337‑TA‑1400, Certain Cameras, Camera Systems, and Accessories Used Therewith (the only ITC case for this patent)

• GoPro complaint filed March 29, 2024; investigation instituted May 6, 2024 (89 FR 37242). Respondents: Arashi Vision Inc. d/b/a Insta360 (Shenzhen) and Arashi Vision (U.S.) LLC d/b/a Insta360 (Irvine, CA).
• '840 was asserted with claims 1–21. By Order No. 9 (Sept. 30, 2024, unreviewed Oct. 25, 2024) and Order No. 24 (Jan. 13, 2025, unreviewed Jan. 31, 2025), GoPro withdrew claims 1–12 and 15–21, leaving only claims 13 and 14 asserted.
• Final Initial Determination (ALJ Doris Johnson Hines, July 10/11, 2025): claims 13 and 14 of the '840 patent not shown to be infringed; claim 14 shown to be invalid; claim 13 not shown invalid; technical and economic prong of domestic industry satisfied for the '840 patent. (Sourced: ITC notice at usitc.gov; reproduction in PTACTS.)
• Commission review (Sept. 11, 2025): reviewed the Final ID findings on limitation 1[g(i)] — "determine a smoothed trajectory of the housing based on a look‑ahead of the trajectory and one or more of a weight‑balance parameter, a low‑light high‑pass parameter, and/or a stickiness parameter" — of claim 1, and the additional limitations of claim 14.
• Commission final determination, February 26, 2026 (91 FR, published Mar. 3, 2026): found no violation of Section 337 as to the '840 patent (and the '052 patent); found a violation only as to U.S. Design Patent No. D789,435; issued a limited exclusion order and cease‑and‑desist order; investigation terminated. Specifically, the Commission reversed the Final ID's finding that the accused cameras satisfy the "low‑light high‑pass parameter" and the additional limitations of claim 14; Insta360 describes this as an added non‑infringement ground on the '840 patent. Trade date was extended to Feb. 17, 2026 due to the federal government shutdown.
• Note on the bond: ALJ recommended 0% bond.

District court — GoPro, Inc. v. Arashi Vision, C.D. Cal. No. 8:24‑cv‑00681 — filed Mar. 29, 2024 asserting six IP rights including the '840 patent; stayed June 3, 2024 under 28 U.S.C. § 1659 at Insta360's request pending the ITC investigation. Per Insta360's Feb. 2026 announcement, the district case resumes following termination of the ITC investigation; final outcome still uncertain.

PTAB — IPR2024‑01164 and IPR2024‑01165 (Petitioner: Arashi Vision (U.S.) LLC d/b/a Insta360), challenging claims 1–21 of the '840 patent. Grounds seen in the record: Ground 1 — Bell (EX1006) and Shi render obvious claims 1–3, 5, 8–11, 14, 16, 18, 19; other references cited include Zhou and Cai (EX1009). Google Patents tags both proceedings "(Final Written Decision)," but I could not verify the FWD outcomes in this session — treat the outcomes as unconfirmed.

Federal Circuit (2026) — Google Patents lists three 2026 CAFC cases in this family: 26‑1392, 26‑1393, and 26‑1752. Given the Feb. 26, 2026 ITC final determination and the 60‑day appeal window noted in Insta360's disclosure, these are consistent with GoPro's appeals of the ITC final determination (and/or related PTAB appeals). I could not verify case captions, parties, or docket entries — do not treat the party identification as authoritative.

5. Explicit uncertainties

1. Claim 1 preamble and claim 16 full text — retrieved only in part. The elements quoted above are verbatim; the transitional/preambular language and claims 4, 7, 12, 17–21 are not verified verbatim and are omitted rather than inferred.
2. CAFC 2026 docket contents — docket numbers 26‑1392, 26‑1393, 26‑1752 come from the Google Patents litigation panel; captions, parties, filing dates and issues were not confirmed against the Federal Circuit's own docket.
3. IPR2024‑01164 / ‑01165 outcomes — only the Google Patents label "Final Written Decision" and petition‑side materials were seen; I did not confirm institution status, final claim dispositions, or any appeal.
4. Precise ITC disposition of claim 14's validity in the final determination — the retrieved Federal Register text was truncated immediately after the Commission's reversal on the low‑light high‑pass finding. Insta360 publicly states the '840 patent claims were held "invalid, not infringed, or both."
5. Date note: you state today is April 26, 2026, while the source page was fetched 2026‑09‑30 (and one Insta360 annual‑report table is dated 2026‑04‑28). I have reported document dates as published and have not attempted to reconcile these.
6. Related family members (same title, overlapping disclosure, useful for a claim‑chart/FTO view): US 10,341,564 (Derbanne et al.), US 11,025,824 (Derbanne et al.), US 10,432,864, plus EP 3854070 A4 and EP 3794813 A1. These are different patents — not the '840 patent.

I'll search multiple litigation databases for this specific patent number.

Let me verify the PTAB and Federal Circuit proceedings.

Litigation Report — U.S. Patent No. 10,958,840 ("Systems and methods for stabilizing videos")

Scope note: The instruction was to search only for litigation involving U.S. Patent No. 10,958,840. All results below were screened so that the asserted patents in each matter include the '840 patent. During searches, related GoPro/Insta360 matters on other patents (e.g., 10,015,413; 10,529,052; 10,574,894; 11,336,832; D789,435) also appeared; I include them only where the '840 patent is also asserted, and I flag them where relevant context overlaps.

Patent identified: US 10,958,840 B2, "Systems and methods for stabilizing videos." Inventors: Cesar Douady, Thomas Derbanne, Maxim Karpushin. Current assignee: GoPro, Inc. Application 16/713,798, filed December 13, 2019; priority date Sept. 19, 2018; granted March 23, 2021; anticipated expiration Apr. 23, 2039. Claim 1 recites determining a smoothed trajectory "based on a look-ahead of the trajectory and one or more of a weight-balance parameter, a low-light high-pass parameter, and/or a stickiness parameter."

Yes — litigation exists. There are three main U.S. forums plus PTAB proceedings, and now appellate activity.

1. ITC Section 337 Investigation

• Matter: In the Matter of Certain Cameras, Camera Systems, and Accessories Used Therewith
• Investigation No.: 337‑TA‑1400
• Plaintiff/Complainant: GoPro, Inc. (San Mateo, CA)
• Respondents/Defendants: Arashi Vision Inc. d/b/a Insta360 (Shenzhen, China) and Arashi Vision (U.S.) LLC d/b/a Insta360 (Irvine, CA) (together, "Insta360")
• Jurisdiction: U.S. International Trade Commission, Washington, D.C.
• Complaint filed: March 29, 2024 (U.S. time); investigation instituted May 6, 2024, 89 Fed. Reg. 37242
• '840 patent role: GoPro asserted claims 1–21 of the '840 patent (along with four other utility patents and design patent D789,435). The investigation was later partially terminated as to claims 1–12 and 15–21 of the '840 patent (Order No. 9, Sept. 30, 2024, unreviewed Oct. 25, 2024), leaving claims 13 and 14 at issue.
• ALJ Final Initial Determination (July 10, 2025, ALJ Doris Johnson Hines):
  • Claims 13 and 14 of the '840 patent have not been shown to be infringed (Findings 5, and the notice's item 5);
  • Claim 14 of the '840 patent has been shown to be invalid (item 11); claim 13 has not been shown invalid (item 12);
  • Domestic industry technical/economical prongs satisfied as to the '840 patent (items 18, 23).
  • (Primary source: USITC notice of issuance of Final ID, https://www.usitc.gov/press_room/documents/notice_of_issuance_final_idrd_1400.pdf)
• Commission review (Sept. 11, 2025): The Commission determined to review, inter alia, the Final ID's findings on the '840 claim 1 limitation "determine a smoothed trajectory … based on a look-ahead of the trajectory and one or more of a weight-balance parameter, a low-light high-pass parameter, and/or a stickiness parameter" (element 1[g(i)]) and the additional limitations of claim 14. It declined to review the no-violation findings for the '894, '832, and '413 patents.
• Commission Final Determination (Feb. 26, 2026; notice published 91 Fed. Reg., FR Doc. 2026‑04156, filed Mar. 2, 2026):
  • The Commission reversed the Final ID's finding that the accused cameras satisfy the "low-light high-pass parameter" and the additional limitations of claim 14 of the '840 patent;
  • Result: no violation of Section 337 as to the '840 patent (no infringement). A limited exclusion order and cease-and-desist order issued only with respect to design patent D789,435 (legacy Insta360 "Ace" models).
  • Status: Investigation terminated; ITC matters do not award damages. GoPro (and Insta360) retain a 60-day right to appeal to the Federal Circuit.
  • (Sources: ITC final determination text at https://www.govinfo.gov/content/pkg/FR-2026-03-03/pdf/2026-04156.pdf; Insta360 press release "Insta360 Secures Complete Victory in ITC Final Ruling"; Morrison Foerster client alert, Feb. 26, 2026.)

2. Parallel District Court Action

• Case: GoPro, Inc. v. Arashi Vision Inc. et al.
• Court / Jurisdiction: U.S. District Court for the Central District of California (Western Division)
• Case No.: 8:24‑cv‑00681
• Plaintiff: GoPro, Inc.
• Defendants: Arashi Vision Inc. d/b/a Insta360; Arashi Vision (U.S.) LLC d/b/a Insta360
• Filed: March 29, 2024
• Judge: Hernan D. Vera; Magistrate Judge John D. Early
• Cause of action: 35 U.S.C. § 271 patent infringement (six U.S. IP rights, including the '840 patent)
• '840 patent role: GoPro alleged the accused Insta360 products implement in-camera stabilization features ("Quick FlowState," "FlowState 2.0") practicing the '840 and '894 patents.
• Status: On Insta360's motion under 28 U.S.C. § 1659 (stay of parallel district court action where a party to the ITC investigation requests a stay), the court stayed the case on June 3, 2024 pending the ITC investigation. One docket-aggregator source lists the case as "Closed," but the parties' own disclosures state the stay will be lifted and the case resumed now that the ITC investigation has concluded; final outcome on the merits is therefore still undetermined. GoPro also reserved the right to move to reopen. (Sources: Docket, https://dockets.justia.com/docket/california/cacdce/8:2024cv00681/[920163](/patent/920163); Insta360/影石创新 public announcement, Feb. 27, 2026.)

3. PTAB Inter Partes Reviews (both on the '840 patent)

Two separate IPRs were filed by Arashi Vision/Insta360 challenging the '840 patent:

A. IPR2024‑01164

• Petitioner: Arashi Vision Inc. (d/b/a Insta360); Patent Owner: GoPro, Inc.; Patent: 10,958,840 B2
• Filed: July 12, 2024; institution granted (Jan. 2025) on grounds that claims were obvious over Bell and Shi
• Final Written Decision: November 25, 2025 — no challenged claims held unpatentable (GoPro prevailed)
• Status: Appealed to the Federal Circuit by Petitioner (Arashi Vision (U.S.) LLC Notice of Appeal filed Jan. 23, 2026)

B. IPR2024‑01165

• Petitioner: Arashi Vision Inc. (d/b/a Insta360); Patent Owner: GoPro, Inc.; Patent: 10,958,840 B2
• Filed: July 12, 2024; institution decision Jan. 22, 2025 (Ground 1: Zhou + Cai, among others)
• Final Written Decision: November 25, 2025 — Judgment: Final Written Decision Determining No Challenged Claims Unpatentable (35 U.S.C. § 318(a)); termination date Nov. 25, 2025
• Status: Final Written Decision — Appealed (Arashi Vision (U.S.) LLC Notice of Appeal filed Jan. 23, 2026)
• (Sources: PTAB oral-hearing transcripts and Board Order granting motions to seal, at https://ptacts.uspto.gov; IP Verse case page https://ipverse.greyb.com/ptab-web/cases/case-details/IPR2024-01165.)

Note: Related IPRs on other GoPro patents (IPR2024‑00984 on 10,015,413; IPR2024‑01052 on 11,336,832; IPR2024‑01186 on 10,574,894) are not '840-patent matters but appeared repeatedly in the same searches.

4. Federal Circuit Appeals (pending)

The Google Patents "Family has litigation" data for US 10,958,840 lists three Court of Appeals for the Federal Circuit dockets:

Caveat (stated explicitly): I could verify the existence and "pending" status of these three CAFC dockets from the Google Patents litigation linkage for the '840 patent, and separately confirmed that Arashi Vision filed a Notice of Appeal on January 23, 2026 in the IPR proceedings. However, I could not independently confirm, from the search results, which specific docket number corresponds to which underlying proceeding, nor the appellant identity for each. The 26‑1392/26‑1393 ↔ IPR and 26‑1752 ↔ ITC mapping above is my inference and should be verified against CAFC docket entries before being relied upon. As of the report date (April 26, 2026), no outcome has been reported for any of the three.

5. Related / adjacent matter (not a U.S. '840 case)

Insta360 (影石创新) and its subsidiary 前海影石 filed a Chinese patent-infringement suit against GoPro-related entities (高途乐贸易(上海)有限公司, 高途乐科技(深圳)有限公司, and a local reseller) in the Jiangsu High People's Court, accepted July 2024, asserting approximately RMB 110,200,000. This is Insta360 asserting its own patents against GoPro in China; it does not involve U.S. Patent 10,958,840 and is noted only for completeness.

Summary of outcomes to date

Bottom line: The '840 patent has been litigated in the ITC, in the Central District of California, and in two PTAB IPRs, with Federal Circuit appeals now pending. Through the ITC's February 26, 2026 final determination, the '840 patent has not been successfully enforced against Insta360 (no violation found), and both IPRs resulted in no claims held unpatentable in the Final Written Decisions. The district court case remains stayed/resumable and its merits are unresolved. If you need docket-level verification of the three CAFC appeals or a PACER-confirmed current status of 8:24‑cv‑00681, those are the two items I'd recommend confirming against the primary sources.

I'll research this now. Note the structured ODP block reports no AIA trials, but the Google Patents record for this patent lists two PTAB matters (IPR2024-01164 and IPR2024-01165) plus Federal Circuit appeals — I need to verify these against live sources.

Two IPRs confirmed. Now let me pin down the claim-level outcomes, panels, and appeals.

Let me verify the IPR2024-01164 outcome and panel, and the appeal dockets.

Proceedings overview

Important data conflict up front: the structured "PTAB proceedings on file" block sourced from the USPTO Open Data Portal (ODP) reports zero AIA trial proceedings against US 10,958,840, while live web sources — including the Google Patents family record for this patent, which carries an explicit "PTAB case IPR2024-01165 filed (Final Written Decision)" and "PTAB case IPR2024-01164 filed (Final Written Decision)" entry, plus third-party PTAB docket aggregators — independently show two IPRs, both filed 2024-07-12, both instituted 2025-01-22, both decided 2025-11-25, and both now on appeal to the Federal Circuit. I am reporting those two proceedings below and flagging them as not coming from the canonical ODP feed, per the instruction to surface proceedings ODP has not indexed; verify against PTAB E2E before relying on either. On the merits as best I can source them: 2 proceedings, 0 claims canceled, 21 of 21 challenged claims sustained; both FWDs appealed by the petitioner; no settlements; no institution denials. Bottom line for a defendant: the PTAB did not invalidate a single claim of this patent — nothing here is a "claims 1-5 are canceled" gift. The patent's only adverse validity finding came from the ITC, not the PTAB (below), and ITC invalidity findings carry no collateral-estoppel effect in district court.

IPR2024-01164 — Arashi Vision Inc. (d/b/a Insta360) v. GoPro, Inc.

• Type: Inter Partes Review
• Filed: 2024-07-12
• Status: Final Written Decision - Appealed (third-party docket verbatim; Google Patents lists it as "filed (Final Written Decision)")
• Judge panel: Not confirmed by the sources I could reach. (The companion '840 IPR, IPR2024-01165, was heard by APJs Brian D. Range, Nabeel U. Khan, and William V. Saindon — do not assume the same panel sat on -01164.)
• Petition grounds: Challenge to claims 1–21 of the '840 patent, presented as Grounds 1 through 7 under 35 U.S.C. § 102 (anticipation) and § 103 (obviousness), over combinations including Bell, Shi, Karpenko, and Sokeila. (Sourced from the Arctic Invent petition-as-filed summary — treat the reference mapping as unverified until you pull the petition from E2E.)
• Institution decision: Instituted 2025-01-22. Reported basis: the Board found a reasonable likelihood that at least one challenged claim was unpatentable (§ 314(a)).
• Final Written Decision: Issued 2025-11-25. Outcome as reported: the Board held that petitioner's prior art failed to show any of the challenged claims unpatentable, maintaining the patent. I could not independently pull the -01164 FWD text and therefore cannot confirm the claim-by-claim disposition (e.g., whether every one of claims 1–21 was separately addressed, whether some claims were not reached, or whether the Board's analysis turned on a claim-construction ruling on the "low-light high-pass parameter"). Do not represent to a court that the FWD expressly sustained any particular claim number without reading the paper.
• Settlement / termination: None. The proceeding ran to FWD; termination date recorded as 2025-11-25.
• Appeal: Yes. A notice of appeal was filed in January 2026 (petitioner-side). The patent's family record lists Federal Circuit appeals 26-1392 and 26-1393; my sources tie 26-1393 to IPR2024-01165, so 26-1392 is the likely companion appeal from -01164, but I could not confirm that docket-to-proceeding mapping and am not asserting it as fact. No CAFC disposition — these appeals are pending as of today.
• Defensive value: This is the broadest unsuccessful attack on the patent — 21 claims, seven grounds, four reference families — and it failed. It is also the strongest available § 325(d) / General Plastic argument against a follow-on petition: the Office has already spent a full trial on this art and sustained the claims.

IPR2024-01165 — Arashi Vision Inc. (d/b/a Insta360) v. GoPro, Inc.

• Type: Inter Partes Review
• Filed: 2024-07-12
• Status: Final Written Decision - Appealed; the docket paper is captioned verbatim "JUDGMENT Final Written Decision Determining No Challenged Claims Unpatentable 35 U.S.C. § 318(a)" (2025-11-25). Third-party outcome tagging: "Patentable."
• Judge panel: Brian D. Range, Nabeel U. Khan, William V. Saindon (APJs).
• Petition grounds: Parallel challenge to the '840 patent filed the same day as -01164. The specific challenged-claim set and art for -01165 are not confirmed by my sources; the oral-hearing transcript shows the arguments focused on claim 1 and claim 14 (including the "low-light high-pass parameter"), with § 103 obviousness as the operative theory.
• Institution decision: Instituted 2025-01-22. The Board also issued a preliminary construction of "low-light high-pass parameter" at institution: a parameter that controls the amount or strength of smoothing using motion blur compensation. Petitioner's counsel argued at the hearing against GoPro's proposed narrowing addition ("by allowing more high frequencies to pass through based on low-light conditions") and relied on claim differentiation against dependent claim 14.
• Final Written Decision: Issued 2025-11-25 — no challenged claims held unpatentable. This is the cleanest documented PTAB win for the patent owner in the docket: the judgment paper itself says "No Challenged Claims Unpatentable." Procedural lead-up: oral argument set 2025-09-15, hearing held (confidential and public transcripts dated 2025-11-04), sealed FWD-related motions granted 2025-12-02. Note that portions of the record — including the confidential hearing transcript and demonstratives — are under seal, so the panel's full reasoning on the sealed secondary-considerations evidence is not public.
• Settlement / termination: None. Termination/judgment 2025-11-25.
• Appeal: Yes — Appeal No. 26-1393 (Federal Circuit), consistent with the family record. Notice of appeal filed 2026-01-23 by Arashi Vision (U.S.) LLC (d/b/a Insta360) as petitioner. Pending; no disposition.
• Defensive value: The judgment is a § 318(a) affirmance of patentability that future petitioners must contend with, and it hands GoPro a favorable institution-stage construction of "low-light high-pass parameter" to carry into district court. But because the FWD sustained the claims, it does not narrow the patent — a new defendant faces the same claim scope GoPro obtained in 2021.

Strategic summary

Claim status on the PTAB record: 0 canceled, 0 narrowed, 21 sustained. Every claim of US 10,958,840 that has been put in front of the PTAB — on the reported -01164 challenge, claims 1–21 — survived. There are no PTAB-canceled claims to point to, no certificate of cancellation, and no disclaimer on the record that I located. The surviving claim set is therefore the full issued set, claims 1–21. The claims that have not been meaningfully tested at the PTAB at claim level in a way I can verify are, effectively, all of them — the FWD outcomes are reported at the whole-proceeding level ("no challenged claims unpatentable"), so I would not treat any individual claim as having been "found patentable" in a way you can quote to a judge.

The one genuine invalidity finding is an ITC finding, not a PTAB finding — and that matters enormously. In Inv. No. 337-TA-1400, GoPro sued Arashi Vision (Insta360) on 2024-03-29. The ALJ (Doris Johnson Hines) issued a Final Initial Determination on 2025-07-10 finding, inter alia, no infringement of the '840 patent. On 2026-02-26 the Commission issued its final determination: no violation as to '840, because claim 13 is not infringed and is invalid, and claim 14 is not infringed and is invalid as anticipated, with the Commission also reversing the ALJ's finding that the accused cameras satisfied the "low-light high-pass parameter" and additional limitations of claim 14. Caveats you must carry: (i) by January 2025 the Commission had terminated the investigation as to claims 1–12 and 15–21 of '840, so only claims 13 and 14 were ever adjudicated at the ITC — the invalidity finding reaches those two claims only; (ii) the invalidity basis for claim 13 is not legible from the published Federal Register notice I relied on; and (iii) ITC patent-invalidity rulings do not have collateral-estoppel effect in district court, so GoPro is not bound by it in the parallel litigation.

Estoppel landscape. § 315(e)(2) estops Insta360 / Arashi Vision (U.S.) LLC and their privies from asserting in the C.D. Cal. case (8:24-cv-00681, GoPro v. Insta360, stayed 2024-06-03 pending the ITC) any ground they raised or reasonably could have raised in IPR2024-01164 and IPR2024-01165 — and those two petitions together swept in claims 1–21 and multiple art families (Bell, Shi, Karpenko, Sokeila), which makes the reasonable-could-have-raised surface wide. For anyone else, no estoppel attaches at all. The full prior-art universe remains open, subject only to § 315(b) (one year from service of an infringement complaint) and the Board's discretionary-denial tools — where § 325(d) (same or substantially the same art already before the Office) and General Plastic / follow-on factors are now live-and-dangerous arguments for GoPro to raise if you are simply re-running Insta360's references. Practical read: fresh art is your path; recycled art is your trap. Sequencing also matters — if you are in a stayed or co-pending district case, expect GoPro to argue the ITC's co-pending § 337 posture and the 2025-11-25 FWDs both cut against discretionary institution.

Pattern signals. (1) Same petitioner, two petitions, same patent, same day — Insta360 filed IPR2024-01164 and IPR2024-01165 on 2024-07-12, and a third against sibling patent 10,574,894 (IPR2024-01186); all three were argued together at a single hearing. That is a coordinated multi-patent campaign keyed to the ITC complaint, not a standalone validity attack — a textbook § 315(b)-deadline-driven filing cluster. (2) Petitioner is appealing both losses (notices of appeal 2026-01-23; Federal Circuit Nos. 26-1393 confirmed, 26-1392 likely the companion), and there is a third CAFC matter, 26-1752, which on timing is almost certainly the appeal from the ITC final determination — GoPro's 60-day appeal window from 2026-02-26 runs to roughly late April 2026, and Insta360 likewise has rights to appeal. So the appellate layer is unsettled, and any FWD-based defense you build today can be unwound on appeal. (3) No defensive aggregator is in this chain. Google Patents renders the PTAB rows under an "Unified Patents PTAB Data" attribution with a blank "Petitioner:" field — that is a data-license attribution, not an identification of Unified Patents as petitioner. The real petitioner of record is Arashi Vision Inc. (d/b/a Insta360), a commercial competitor. Do not brief this as an aggregator-funded challenge.

Recommended next steps

• Pull the two FWDs from PTAB E2E before you rely on anything above. The -01165 judgment paper ("Determining No Challenged Claims Unpatentable," 2025-11-25) is the ground truth for the § 318(a) disposition; the -01164 FWD is the one I could not independently verify. Download both, plus the 2025-01-22 institution decisions (which contain the "low-light high-pass parameter" preliminary construction) and the 2025-11-04 hearing transcripts — start at https://ptacts.uspto.gov/ptacts/ (PTAB E2E / PTACTS).
• Do not plead "the PTAB canceled the claims." It did not. The accurate defensive framing is: claims 1–21 were challenged in IPR2024-01164/01165 and sustained; claims 13 and 14 were separately held invalid by the ITC in 337-TA-1400; appeals are pending. If your demand letter or invalidity contention quotes the ITC invalidity finding, cite it as an ITC determination with no collateral-estoppel effect, quote the Federal Register notice verbatim (https://www.govinfo.gov/content/pkg/FR-2026-03-03/pdf/2026-04173.pdf — "no violation for '840 patent because claim 13 is not infringed and is invalid claim 14 is not infringed and is invalid as anticipated"), and do not impute it to the Board.
• Track the three Federal Circuit appeals — 26-1392, 26-1393, and 26-1752 (per the Google Patents family record: https://portal.unifiedpatents.com/litigation/Court%20of%20Appeals%20for%20the%20Federal%20Circuit/case/26-1392, .../26-1393, .../26-1752). Confirm docket-to-proceeding mapping and claim-scope issues on CourtListener before briefing. If the -01164/-01165 appeals are affirmed, GoPro's § 325(d) hand gets stronger; if remanded or reversed, an invalidity theory may reopen.
• If you are a defendant being newly asserted against, calendar the § 315(b) one-year bar from your service date now, and budget for a petition built on art outside the Bell / Shi / Karpenko / Sokeila families — that is the only ground set that is both non-estopped (you are not an Insta360 privy) and not primed for § 325(d) discretionary denial.
• Watch the C.D. Cal. case, GoPro, Inc. v. Insta360, No. 8:24-cv-00681 (C.D. Cal.), which was stayed on 2024-06-03 pending the ITC investigation. Now that the ITC investigation terminated in its entirety on 2026-02-26, expect a motion to lift the stay — and check how Judge-tied claim constructions of "look-ahead," "weight-balance parameter," "low-light high-pass parameter," and "stickiness parameter" from the ITC and the two institution decisions land in the district court.
• Gap-check the ODP discrepancy explicitly. If you are relying on the structured ODP block in a filing or opinion, note that it returned zero AIA proceedings while the PTAB docket and the patent's own family record show two. That kind of mismatch is exactly what opposing counsel will probe — reconcile it against PTAB E2E and the PTABIS API rather than assuming either source is wrong.

I'll research the assignment record for this patent before assembling the report.

Assignment & Ownership Reconstruction — US 10,958,840 B2

Patent: US 10,958,840 B2 — "Systems and methods for stabilizing videos"
App. No.: 16/713,798 (continuation of 16/587,811, filed 2019‑09‑30; priority chain traced to 2018‑09‑19)
Filed: 2019‑12‑13 · Granted: 2021‑03‑23 · Anticipated expiration: 2039‑04‑23
Assignee of record: GoPro, Inc. (Delaware), 3025 Clearview Way, San Mateo, CA

Data-provenance note (read before using the reel/frame column): The ownership events and execution/recordation dates below are taken from the Google Patents legal‑events record for US10958840B2 (https://patents.google.com/patent/US10958840/en) and cross‑checked against the Assignment Center / Assignment Search listings. Google Patents surfaces these events without reel/frame numbers, and the per‑patent assignment abstract at assignmentcenter.uspto.gov was not retrievable during this session. I have therefore marked every reel/frame as NOT RETRIEVED rather than invent values. Verify each entry at https://assignment.uspto.gov/patent/index.html (search "10958840") or https://assignmentcenter.uspto.gov/ before relying on reel/frame citations. This is a single material gap in an otherwise complete chain.

Inventors

Pattern check — mass inventor departure within 12 months of filing: NOT PRESENT. All three inventors continue to appear as named inventors on later GoPro continuations from the same family, e.g. US 11,228,712 B2 (granted 2022‑01‑18), US 12,256,147 B2 (filed 2023‑06‑30) and US 12,211,520 B2 (filed 2023‑08‑03). There is no evidence of the entire inventorship group leaving GoPro and no evidence of an inventor‑driven portfolio transfer. The patent carries a terminal disclaimer (noted on its face), tying it to a family sharing a common prior‑art date — relevant because any encumbrance or sale of the '840 almost certainly has to travel with its siblings.

Original assignee

GoPro, Inc. — the entity named on the issued patent as assignee, and the entity that received the inventor assignment executed/recorded on 2019‑12‑13.

• Primary line of business: consumer action cameras, camera systems, mounts/accessories, and video‑editing software (HERO and MAX product lines).
• Does it ship a product embodying the claims? Yes, on the record. GoPro's ITC complaint in Inv. No. 337‑TA‑1400 and its public domestic‑industry charting (Petitioner/Complainant Exhibit 2006, "Public Version of U.S. Patent No. 10,958,840 Domestic Industry Chart," filed in IPR2025‑00017) chart GoPro's own HERO/MAX HyperSmooth stabilization against the '840 claims. GoPro also asserts it "owns by assignment all rights, title, and interest" in the asserted patents, including the '840 (C.D. Cal. complaint, Case No. 8:24‑cv‑00681).
• Current status: Operating. Publicly traded (NASDAQ: GPRO), Delaware corporation. Not in bankruptcy and not dissolved. It is, however, financially leveraged: on 2025‑08‑04 GoPro granted two separate security interests in its patent portfolio — one to Farallon Capital Management, L.L.C., as Agent and one to Wells Fargo Bank, National Association, as Agent — which is the classic structure for a term‑loan collateral agent plus a revolver/ABL collateral agent. That is a financing event, not a sale, but it is the single most important forward‑looking watch item on this asset.

Assignment timeline

All reel/frame values NOT RETRIEVED — see provenance note. Dates are executed/recorded as reflected in the Google Patents legal‑events record.

• 2019‑12‑13 (executed) / recorded 2019‑12‑13 — Reel NOT RETRIEVED

  • Conveyance: Assignment ("Assignment of Assignors' Interest")
  • Assignor: Cesar Douady; Thomas Derbanne; Maxim Karpushin (jointly)
  • Assignee: GoPro, Inc.
  • Correspondent: NOT RETRIEVED. No recurring‑correspondent flag can be supported from this record — flag left open.
  • Context: Original employment assignment — inventors convey title to GoPro on filing of the continuation.

• 2020‑10‑19 (recorded) — Reel NOT RETRIEVED

  • Conveyance: Security Agreement (Security Interest)
  • Assignor: GoPro, Inc.
  • Assignee: JPMorgan Chase Bank, N.A., as Administrative Agent
  • Correspondent: NOT RETRIEVED for this patent. Sibling‑record context: GoPro's IP collateral recordings on the JPMorgan credit facility were filed through corporate‑services filer Cogency Global Inc., 1025 Vermont Ave NW, Suite 1130, Washington DC 20005 (e.g. USPTO record Reel 6348/0594, GoPro → JPMorgan, executed 2018‑03‑26, recorded 2018‑06‑07, per the TTAB assignment abstract of title). This is a registered‑agent/corporate‑services filer handling blanket lien perfection — not an NPE attorney, so it does not trip the repeat‑correspondent troll signal.
  • Context: Securitization — blanket security interest over GoPro's patent portfolio under its secured credit facility; no transfer of title.

• 2021‑01‑25 (recorded) — Reel NOT RETRIEVED

  • Conveyance: Release (Release of Patent Security Interest)
  • Assignor: JPMorgan Chase Bank, N.A., as Administrative Agent
  • Assignee: GoPro, Inc.
  • Correspondent: NOT RETRIEVED (same corporate‑services pattern as above).
  • Context: Lien release — GoPro's JPMorgan patent collateral released; title was never transferred at any point, so this simply clears the encumbrance record.

• 2025‑08‑04 (recorded) — Reel NOT RETRIEVED

  • Conveyance: Security Agreement (Security Interest)
  • Assignor: GoPro, Inc.
  • Assignee: Farallon Capital Management, L.L.C., as Agent
  • Correspondent: NOT RETRIEVED.
  • Context: Securitization — new 2025 secured financing; collateral agent for GoPro's lenders.

• 2025‑08‑04 (recorded) — Reel NOT RETRIEVED

  • Conveyance: Security Agreement (Security Interest)
  • Assignor: GoPro, Inc.
  • Assignee: Wells Fargo Bank, National Association, as Agent
  • Correspondent: NOT RETRIEVED.
  • Context: Securitization — companion lien recorded the same day as the Farallon lien, consistent with a multi‑tranche credit facility (Farallon and Wells Fargo acting for different lender groups). GoPro remains record owner throughout.

Bottom line on the chain: there is exactly one conveyance of title in the entire history (inventors → GoPro, 2019). Everything recorded afterwards is a collateral lien or a lien release. GoPro, Inc. is and always has been the owner of record.

Timeline diagram

timeline
    title Ownership of US 10958840
    2018 : Priority application filed
    2019 : Continuation filed Dec 13
         : Inventors assign to GoPro Inc
    2020 : JPMorgan security interest recorded
    2021 : Patent issued Mar 23
         : JPMorgan security interest released
    2024 : GoPro files ITC case vs Insta360
    2025 : Farallon and Wells Fargo liens recorded
    2026 : ITC final ruling on the 840 patent

Litigation overlay (needed to read the NPE signals correctly)

• 2024‑03‑29 — GoPro files ITC Section 337 complaint against Arashi Vision Inc. d/b/a Insta360 and Arashi Vision (U.S.) LLC; instituted as Inv. No. 337‑TA‑1400 (89 Fed. Reg. 37242) asserting the '413, '052, '894, '840, '832 and D'435 patents. GoPro is the complainant in its own name.
• 2024 — Parallel district court action, Arashi Vision v. GoPro / GoPro v. Arashi Vision, C.D. Cal. No. 8:24‑cv‑00681, stayed 2024‑06‑03 pending the ITC case.
• 2024‑07‑12 — Insta360 petitions for IPR: IPR2024‑01164 and IPR2024‑01165 (the latter on the '840; instituted 2025‑01‑22).
• 2026‑02‑26 — ITC final determination: no violation as to all five asserted utility patents, including the '840 (non‑infringement plus invalidity/partial invalidity); limited exclusion order issued only on the design patent D'435 and only as to legacy Insta360 Ace models. GoPro's appeals are docketed at the Federal Circuit (26‑1392, 26‑1393, 26‑1752).
• No assignment recorded in the six months before the March 2024 complaint. The nearest prior recorded events are the January 2021 lien release and the December 2019 inventor assignment.

NPE / troll-pattern signals

1. Shell-entity transfer — NOT PRESENT. The chain contains no licensing-only LLC. The only title transfer is inventors → GoPro (2019‑12‑13). No "IP / Holdings / Licensing / Ventures" assignee appears anywhere in the record; no registered‑agent address, no single‑purpose Delaware or Texas LLC.

2. Known asserter in the chain — NOT PRESENT. None of Acacia, Marathon, Intellectual Ventures, IPNav, Wi‑LAN, Conversant/Mosaid, Vringo, Pendrell, Innovatio, MPHJ, Lumen View, Round Rock, Document Generation Corp or any Spangenberg‑linked entity appears as assignor or assignee. The counterparties on the recorded events are GoPro, Inc., JPMorgan Chase Bank N.A., Farallon Capital Management L.L.C. and Wells Fargo Bank N.A. — two diversified banks and one investment manager, all in secured‑lender capacities, none of them patent asserters.

3. Repeat correspondent across the chain — UNCLEAR (data gap, leaning not present). I could not retrieve the correspondent of record for any of the '840 recordings, so I cannot state a recurrence. What can be said: the family's comparable GoPro collateral recordings were filed by the corporate‑services firm Cogency Global Inc. (sibling record Reel 6348/0594). A repeat corporate-services filer on lien perfection is the expected, benign pattern and is not an NPE tell — the signal is recurrence of a litigation attorney, and no such name appears in the retrievable records. Do not treat this entry as a finding either way without pulling the actual '840 reel/frame abstracts.

4. Cascading transfers — NOT PRESENT. One title transfer in 2019, then nothing but liens and a release. No chained LLCs, no transfers inside 24 months, no shared correspondent addresses among successive assignees. The two 2025‑08‑04 lien records are same‑day collateral grants to lenders of the same grantor — the opposite of a cascading assignment chain.

5. Pre-litigation transfer — NOT PRESENT. The first infringement suit naming the '840 is the March 2024 ITC complaint (and the 2024 district court case). The last change of any kind on the record was the 2021‑01‑25 lien release — roughly 38 months earlier. No assignment, no shell conveyance, no standing‑cleaning transfer sits in the six‑month window.

6. Bankruptcy fire‑sale — NOT PRESENT (watch item). GoPro has not filed Chapter 7 or 11 and no patent sale in a bankruptcy proceeding has occurred. However, the paired 2025‑08‑04 security interests to Farallon Capital Management and Wells Fargo — both recorded as collateral agent, both dated the same day — indicate GoPro has pledged its patent portfolio to secured lenders at a time of continued operating losses. If GoPro were to default or restructure, this patent portfolio, including the '840 and its terminal‑disclaimer‑linked family, would be the collateral at the centre of any 363 sale. That is a forward risk, not a present finding.

7. Privateering — NOT PRESENT. GoPro litigates in its own name against a direct competitor (Insta360) and has never transferred the '840 to a third party to assert on its behalf. No SEC filing, EFF or Patent Progress coverage indicates a proxy assertion vehicle here.

8. Defensive aggregator — NOT PRESENT. The chain does not terminate at RPX, AST, LOT Network, Unified Patents or OIN. Note that Unified Patents appears in this patent's record only as the publisher of litigation/PTAB metadata on the Google Patents page, and as the likely filer of the IPR challenge docket entries — it is a challenger here, not an owner.

Verdict

Operating-company assertion.

GoPro, Inc. has been the record owner since the sole inventor‑to‑assignee conveyance of 2019‑12‑13, and every subsequent recorded event is a secured‑lender collateral lien (JPMorgan recorded 2020‑10‑19 and released 2021‑01‑25; Farallon Capital Management and Wells Fargo both as Agent recorded 2025‑08‑04) rather than a transfer of title. GoPro ships products (HERO/MAX with HyperSmooth) that its own ITC domestic‑industry charting maps onto the '840 claims, and it asserted the patent in its own name against a real competitor, Insta360, in ITC Inv. No. 337‑TA‑1400 (filed 2024‑03‑29) and C.D. Cal. No. 8:24‑cv‑00681 — losing on the '840 at the ITC on 2026‑02‑26, with Federal Circuit appeals (26‑1392/26‑1393/26‑1752) pending. No shell entity, no listed NPE, no cascading LLC transfers, and no transfer inside six months of suit appear anywhere in the chain.

Confidence limits: high on the identity and sequence of ownership events; medium‑low on reel/frame citations, because the Assignment Center per‑patent abstract was not retrievable in this session and I will not reconstruct reel/frame numbers from inference. Pull the reel/frame index — and in particular the correspondent of record for each of the five entries above — at https://assignment.uspto.gov/patent/index.html (search "10958840") before filing anything that depends on them. The only genuinely notable feature of this chain is not an NPE pattern but a solvency signal: two same‑day collateral grants to Farallon and Wells Fargo in August 2025 over a portfolio whose crown‑jewel stabilization patents are now being invalidated or narrowed in litigation.

I'll research the prior art of record for US 10,958,840, starting with the USPTO/Google Patents citation data.

Let me pull the full "References Cited" list and the Justia/FPO record for the '840 patent.

Let me dig into the IPR grounds and the specific prior-art references (Bell, Shi, Karpenko, Sokeila, Zhou, Cai, Kwatra).

Prior-Art Analysis — U.S. Patent No. 10,958,840 B2

"Systems and methods for stabilizing videos" — GoPro, Inc. (Douady, Derbanne, Karpushin)
App. 16/713,798 · filed 2019-12-13 · priority 2018-09-19 · granted 2021-03-23 · 21 claims

Sources used (all searched live this session): the granted-patent PDF at patentimages.storage.googleapis.com/.../US10958840.pdf (the "(56) References Cited" front page), the Google Patents page patents.google.com/patent/US10958840B2/en, the Justia/FPO records for the '840 and its siblings, the EPO PISE search report for the EP family member, the PTACTS IPR petition/declaration/transcript dockets for IPR2024‑01164 and IPR2024‑01165, and the ITC record in 337‑TA‑1400 (docketalarm.com/cases/International_Trade_Commission/337-1400/).

0. Important framing before the tables

Three things must be said up front, because they change how the list below should be read.

1. "Cited on the face" ≠ "anticipates." The front-page citations under 35 U.S.C. § 102/§ 103 are the Examiner's and applicant's search results. The '840 patent issued over all of them — none produced a claim rejection that survived. Listing a reference under § 102 below means only that it is of the right statutory category and date to be § 102 art, not that it discloses every limitation.
2. The operative prior art against this patent was run as § 103 combinations, not § 102. In IPR2024‑01164 the petitioner's grounds are expressly "renders obvious" grounds: Ground 1 (Bell + Shi), Ground 2 (Bell + Shi + Karpenko, claims 4, 20), Ground 3 (Bell + Shi + Bell Paper, claims 6, 7, 21), Ground 4 (Bell + Shi + Sokeila, claim 12), and further grounds for the remaining claims. The arcticinvent docket summary describes the petition as challenging "21 specific claims based on anticipation (102) and obviousness (103)," but the substantive argument reproduced in the petition TOC is § 103. Both IPRs ended 11/25/2025 with no challenged claims held unpatentable (GoPro prevailed), and both were appealed 2026-01-23.
3. § 102 requires a single reference disclosing every limitation. For independent claim 1 that means, at minimum: housing + optical element + image sensor + position sensor; a trajectory; a smoothed trajectory determined based on a look‑ahead and one or more of a weight‑balance parameter, a low‑light high‑pass parameter, and/or a stickiness parameter; a temporal horizon of motion … enabling approximation of intentional motion … while causing a delay; and the viewing‑window/punchout/selective‑cropping output recitation. No reference of record is a clean single‑reference § 102 anticipation of claim 1 or claim 16. The closest candidates are identified below with the specific element each fails.

1. U.S. patent documents cited on the face of the '840 (Category: § 102(a)(1) / § 102(a)(2) art)

Foreign patent documents on the face (all § 102(a)(1)/(a)(2) category, none individually anticipatory):

Non-patent literature on the face:

Verification caveat: the NPL set above is reproduced from the '840-family record; I could not confirm line-by-line that every NPL item (particularly the Kamali/Kopf/O'Donovan entries) appears on the printed face of the '840 itself rather than being inherited from its sibling filings (US 11,025,824 / US 11,228,712 / US 11,696,027). The Berkeley "Spatial Audio Lecture Series" entry and the PCT search-report entries are confirmed on the '840-family record.

2. The prior art actually asserted against the '840 (the operative art)

This is the set that matters for validity risk. It overlaps only partly with the face citations.

The EPO search report for the EP family member (EP 3854070 A4) categorizes every one of the Bell, Wu, Beysserie, Cai, EP 2805482 and Jenny references as "[A] — general state of the art, not considered of particular relevance," and lists EP 3794813 A1 (GoPro) as an "[E]" (earlier application, published on/after the filing date). That is independent corroboration that the family was considered novel over these documents.

3. § 102 verdict by claim — what could potentially anticipate

4. Contradictions and verification flags

1. Contradiction with the prior litigation report. The earlier section states that the ITC Final ID found claim 13 "not shown invalid." Insta360's own February 27, 2026 disclosure (影石创新 / 9fzt / sina / 10jqka, all published 2026‑02‑27) states that for the '840 the ITC confirmed the accused products "不构成侵权且专利权利要求无效/部分无效" — i.e., non-infringement and the claims invalid or partially invalid. These cannot both be right as to claim 13's validity, and the Federal Register text I could reach was truncated before the claim-14 validity disposition. Treat the final validity status of individual '840 claims as unresolved pending the full Commission opinion. (Note the earlier section also flagged this as uncertainty #4 — the new Chinese-language sources deepen the conflict rather than resolve it.)
2. Sokeila — I could not verify a publication number or date for the reference the petitioner cites as EX1008. Do not populate a number.
3. Zhou (US 10,911,677 B2) issued 2021‑02‑02, after the '840 effective filing date; it is only § 102(a)(2) art if its effective filing date precedes 2018‑09‑19. Unverified.
4. OCR ambiguity on the printed face. The '840 PDF's "References Cited" column returns alternate renderings for at least five entries: US 8,022,948/8,022,048; US 9,426,430/9,426,260; US 10,284,794/10,284,754; US 10,542,215/10,542,115/10,542,221; "Derbaane"/"Derbanne"; "Enoch"/"Endoh"; "Pea"/"Moriya". I have not auto-corrected any of them.
5. Face-citation list is not complete. The retrieved front page was partially truncated (the "U.S. PATENT DOCUMENTS" block shows a "(Continued)" marker and a second column). Additional entries almost certainly exist — e.g., the sibling US 11,228,712 face lists US 4,864,409 A (Platte) and US 7,623,153 B2 (Hatanaka), and the US 11,025,824 face renders "10,441,564" for "10,341,564." Treat Tables 1 and 2 as the verified subset, not the complete 37-odd references of record.
6. Which reference invalidated claim 14 at the ITC was not identified in any source I could reach; the ITC's invalidity ground for the '840 (the docketalarm ID excerpt shows the '840 section but only the non-infringement and claim-construction subsections) remains unverified. Given the ITC's '052 case used Okubo and Sokeila, those are plausible candidates, but that is inference, not evidence.

5. Bottom line

• The face-of-patent art is a broad, mostly peripheral stabilization/omnidirectional-capture set (Gilbert, Kawahara, Garbow, Hollinger ×2, Davis, Van Hoff, Grundmann US 9,374,532, Aguilar, Beysserie US 9,787,902, Miyajima, Cho, Gilmour, Smolyanskiy, Francois, Kunishige) plus three WO publications and a small NPL set, and — notably — two same-family GoPro references (US 10,341,564 Derbanne and US 10,432,864 Douady) that are protected by the § 102(b)(2)(C) common-ownership exception.
• No cited reference is a clean single-reference § 102 anticipation of claim 1 or claim 16, because the claims' distinguishing recitations — the look‑ahead plus the weight‑balance / low‑light high‑pass / stickiness parameter, plus the temporal-horizon-causing-a-delay limitation — are not found together in any one document.
• The best § 102 candidates are in the asserted art, not the face art: Karpenko US 2014/0160309 A1 (claims 3, 4, 6, 7, 20, 21 — iterative coarse-to-fine smoothing with a constraint/projection step) and Bell US 2017/0332018 A1 (claim 1/16 look-ahead smoothing and crop-polygon punchout). Shi US 10,462,370 B2 is the correct § 102(a)(2) formulation for the low-light high-pass limitation, but that theory failed in the IPR.
• Empirically, the prior art has not defeated this patent: both IPR Final Written Decisions (11/25/2025) held no challenged claims unpatentable, the ITC found no violation on the '840 (final determination 2026‑02‑26), and the EPO search report graded the closest art as merely category "[A]." The live risk on the '840 today sits in the pending Federal Circuit appeals (26‑1392, 26‑1393, 26‑1752) and the revived C.D. Cal. case 8:24‑cv‑00681, not in any single reference on the patent's face.

I'll ground this analysis in the actual prior-art references from the PTAB record rather than speculating. Let me pull the petition grounds and the references themselves.

Good — I have the ground structures. Let me pin down the reference identities and the remaining grounds.

Let me identify the exact references and pull the Board's reasons for rejecting the combinations.

Obviousness Analysis — U.S. Patent No. 10,958,840 B2

"Systems and methods for stabilizing videos" (GoPro, Inc.) — 35 U.S.C. § 103

Scope note. This analysis is built on (a) the authoritative patent text in-context, and (b) the actual prior-art record identified for this patent in the PTAB petitions (IPR2024‑01164, IPR2024‑01165), the institution decisions, the oral-hearing transcripts, and the patent-owner responses. Per the operating rules, I do not auto-correct any identifier; where a reference's identity is uncertain I say so rather than substituting a "corrected" number. Where search results and my own reconstruction diverge, I flag the divergence.

⚠️ Critical framing up front: the combinations below are the ones that were actually advanced and litigated. They are not hypothetical. Both IPRs were instituted on those grounds, and both ended in Final Written Decisions on November 25, 2025 holding that no challenged claim was unpatentable. An obviousness analysis that stopped at "a POSITA would have been motivated to combine" would therefore be materially incomplete. Section 7 explains how the affirmative case was defeated.

1. The target claim, decomposed

The patent's own "prior art" metadata on the source page reads: Prior art keywords: trajectory, image capture, visual content, smoothed, capture device; Prior art date: 2018‑09‑19. The claims, however, are narrower than those keywords suggest — the novelty hook is not "smooth a trajectory," it is which parameters modulate the smoothing and the look-ahead temporal horizon.

Claim 1 (independent, "image capture system") decomposes as:

Claim 16 is the method mirror. Dependents 13, 14 and 15 respectively add the weight‑balance, low‑light high‑pass, and stickiness parameters. Claims 6–7 add the minimization/iterative-projection machinery; 11–12 add buffering (and specifically a circular buffer); 8–10 add the asymmetry (past weighted less than future). (Claims 4, 7, 12 and 17–21 were not retrieved verbatim in the earlier session; the content attributed to them here is drawn from the petitions' own quotation of the claim language, which is quoted inline below with the sentence text visible in the PTACTS excerpts.)

The single most important observation for §103: the specification itself states that the objective function "consist[s] of groups of terms… one term expresses the L2‑norm of interframe angular velocity and the other term expresses the angular acceleration," weighted by constants. That is a textbook, decades-old trajectory-smoothing cost function. The inventive weight therefore rests almost entirely on (a) the named parameterization and (b) the causal structure (look-ahead → delay).

2. The prior art actually relied on

3. The combinations asserted, and the element mapping

3.1 Combination A — Bell + Shi (IPR2024‑01164, Ground 1; asserted against claims 1, 13, 14, 15, 16 and the dependents)

Claim 1 mapping (petitioner's theory):

3.2 Combinations B–D — Bell + Shi plus a secondary reference (IPR2024‑01164, Grounds 2–4)

• Ground 2 — Bell + Shi + Karpenko, claims 4 and 20: claim 4 requires "iterations of smoothing including a fine scale approach to smooth on a small scale and a coarse approach to smooth on a large scale, the iterations of smoothing resulting in convergence." Karpenko supplies this verbatim in substance ("Running coarse to fine may permit [the algorithm] to be run for significantly fewer iterations at each step"; improves the "rate of convergence").
• Ground 3 — Bell + Shi + Bell Paper, claims 6, 7 and 21: claim 6 requires "minimization of a value, a score, or a metric that expresses an amount of high frequencies while respecting a crop constraint"; claim 7 requires iterative minimization with an update step and a projection step that re‑projects the estimate when constraints are not satisfied. The Bell Paper's non-linear gyroscope filter supplies the high-frequency objective; the crop constraint is supplied by Bell's polygon-within-image limitation.
• Ground 4 — Bell + Shi + Sokeila, claim 12: Sokeila supplies the "circular buffer" (ring buffer) for storing frames to be stabilized.

Grounds 5–7 of IPR2024‑01164 were not fully retrieved in this session; based on the numbering pattern they addressed the remaining dependent claims (e.g., 8–10, 13, 15, 17–19). ⚠️ I will not speculate as to their content.

3.3 Combination E — Zhou + Cai (IPR2024‑01165, Ground 1; asserted against claims 1–3, 5–11, 14–16, 18, 19, 21)

This is the cleanest structural mapping because Zhou is a self-contained image-capture-device reference:

3.4 Combinations F–H — Zhou + Cai plus a secondary reference (IPR2024‑01165, Grounds 2–4)

• Ground 2 — Zhou + Cai + Karpenko, claims 4, 20: same coarse-to-fine / convergence rationale as in 3.2.
• Ground 3 — Zhou + Cai + Sokeila, claim 12: circular/ring buffer; petitioner argued "there are finite types of buffers for storing visual content during video stabilization," making the selection an obvious design choice from a known, finite set, with predictable results (FIFO; no shifting operations).
• Ground 4 — Zhou + Cai + Bell Paper, claims 13 and 17: the Bell Paper's minimization of rotational velocity and acceleration supplies the claimed adjustable weight‑balance between those two minimization terms.

4. Why a POSITA would have been motivated to combine — the articulated rationales

The petitions relied on recognizable KSR rationales rather than mere adjacency of subject matter:

1. Same field, same problem, same solution architecture. Every primary reference is directed to digital video stabilization by smoothing a device-motion trajectory and cropping/punching out a stabilized view. Zhou and Cai are both expressly "directed to… the stabilization of videos" (Zhou Abstract; Cai Abstract), and both are concerned with distinguishing intentional camera motion from jitter. Bell, the Bell Paper, Kwatra, Karpenko and Sokeila are all gyroscope/IMU-driven trajectory-smoothing approaches.
2. Known technique applied to a known device ready for improvement (KSR factor). Petitioner's argument was that "incorporating Cai's known technique for distinguishing intentional motion from jitter would have been an obvious way to enhance the accuracy and quality of the stabilization taught by Zhou" — i.e., Zhou already analyzed future-frame movement, and Cai supplies a bounded, quantified future-frame window (five frames) that makes that analysis threshold-driven.
3. Predictable results from combining familiar elements. The secondary-reference grounds (Karpenko, Sokeila) are classic "known technique / design choice" arguments: coarse-to-fine iteration "improves the rate of convergence," and a ring buffer is a well-known FIFO structure requiring no shifting operations. Petitioner argued that "[a]pplying a known component (a circular buffer) to a known system to perform its conventional function (storing frames) would yield predictable results."
4. Obvious to try from a finite set. For the buffer-type limitation: "finite types of buffers… including linear buffer and ring buffer."
5. Design incentive / market forces. Both Bell and the '840 patent address mobile/handheld capture, where processing power and battery consumption are the binding constraints — the same design pressure that motivates look-ahead-buffered smoothing and coarse-to-fine iteration. Petitioner separately argued the position information is derived from the inertial sensor (not from the imagery) precisely because doing so is "computationally cheaper."
6. The claimed objective function is the prior art's objective function. The specification's cost (L2 norm of interframe angular velocity + angular acceleration, weighted) is the same functional form as Kwatra's path functional and the Bell Paper's non-linear filter. Petitioner leveraged this hard in the related '894 matter, arguing "there's no dispute that Kwatra is about minimizing a combination of both velocity and acceleration."
7. The parameterization maps onto adjustable, disclosed structures. For "stickiness," petitioner argued Zhou's Gaussian weighting is adjustable because "by just changing the standard deviation of the Gaussian curve… you can have different weighting functions which will affect… how much you consider to follow the past frames."
8. One reference's silence invites the other. Where Zhou "does not discuss what type of buffer is used," petitioner argued a POSITA "would be motivated to turn to prior art, such as Sokeila."

5. The claim-construction predicates that decide the case

Two constructions were outcome-determinative, and both are worth stating plainly because they are the §103 battleground:

(a) "low‑light high‑pass parameter." The Board at institution adopted Insta360's functional construction, treating the term as the applicant's coined term and construing it as "a parameter that controls the amount or strength of smoothing used in motion blur compensation" — essentially synonymous with a motion-blur correction parameter. GoPro's proposed construction required the parameter to (i) be based on low-light conditions and (ii) operate by allowing more high frequencies to pass. Petitioner's rebuttal was textual: the specification says the parameter "controls the amount/strength of smoothing performed to generate the smoothed trajectory," that 0 = "full smoothing… no motion blur compensation" and 1 = "full motion blur compensation," and that low light is merely one condition that produces long exposure time — "Nowhere does the specification limit this parameter to address only blur caused by low light." Under the broad construction, Shi's blur-tracking virtual-camera adjustment reads on the limitation. Under GoPro's narrow construction, it does not.

(b) "temporal horizon of motion… enabling approximation of intentional motion." At institution the Board agreed with petitioner that Cai's five-frame window characterizes "zero motion" as "jitter or shake" and non-zero motion as "an intentional slow movement of the camera," and that looking "some number of frames into the future" is a temporal horizon in the claim's sense. Patent Owner's position was that "temporal horizon involves assessing camera movement patterns over time to interpret the trajectory as either intentional or unintentional," and that Cai merely discriminates motion, it doesn't approximate intent.

6. Graham factors and secondary considerations

• Scope/content of prior art and differences: as tabulated above. The differences the '840 patent can fairly claim are (i) the specific named trio of tuning parameters, and (ii) the express causal requirement that the look-ahead horizon causes a delay. Neither is a structural difference; both are characterization/parameterization differences over the same smoothing pipeline.
• Level of ordinary skill: the petitions used a POSITA with "a bachelor's degree in engineering or computer science, or equivalent field" plus relevant experience — a routine level given the art is dense, incremental, and heavily patented (the justia citation list for this family alone runs to ~90 patent references).
• Objective indicia (GoPro's case, Ex. 2001 ¶¶71–81): HyperSmooth as the practice of the '840 claims; 50M+ action cameras sold since 2009 with the HERO series featuring HyperSmooth "at the forefront"; a leading U.S. action-camera share "larger than all competitors combined"; brand partnerships (NHL, Red Bull, Adobe); CES Innovation Award 2019 and a Technology & Engineering Emmy (2022) for "In Camera Sensor and Software Stabilization" (GoPro and Socionext). GoPro's nexus theory tied the success to "specific features of GoPro's patented technologies, such as its innovative use of specific stabilization parameters, including the weight‑balance parameter, low‑light high‑pass parameter, and stickiness parameter." Note the licensing/attribution point: the Emmy expressly credits GoPro and Socionext jointly, which a challenger would use to attack nexus.
• The counter-indicia petitioner raised as "teaching away" / inoperability: GoPro's expert (Dr. Foroosh) argued that combining Shi's reactive blur detection and estimation with Bell's anticipatory stabilization "would undermine Bell's look‑ahead functionality," because "Bell is focused on pre-emptively adjusting for future motion, while Shi reacts only after blur is detected," and that the combination "would thus render Bell's system inoperable for its intended purpose." GoPro also argued the references are limited to "cell phones and other mobile devices for recording casual video," whereas the '840 patent is directed to more demanding applications ("bodycams in fast sports, mobile agile platforms").

7. What the adjudicated record actually says — and why the affirmative case failed

This is the part that a purely doctrinal §103 write-up would omit, and it matters because it is currently the ground truth for this patent.

A. Both IPRs were instituted, and both failed on the merits.

⚠️ Flags: Google Patents labels both proceedings "(Final Written Decision)," the ipverse case pages list "Status: Final Written Decision – Appealed, Termination Date Nov 25, 2025," and an independent Chinese-language report dated Nov. 25, 2025 states that the USPTO's first final decision in the Insta360-vs-GoPro IPR series held that "the prior evidence [Insta360] adduced fails to prove that either petition renders GoPro's claims invalid," thereby upholding the patent's validity. Three independent sources agree on the outcome and date. I have not retrieved the FWD opinion text itself, so I cannot quote the Board's precise reasoning; the reasoning below is reconstructed from the parties' briefs and the oral-hearing transcripts, which is a lower-quality source. Treat it accordingly.

B. Why the Bell + Shi case failed (from the record as it stood at the hearing). The decisive themes in the record were:

1. No articulated motivation beyond "same field." GoPro's expert characterized petitioner's reasonable-expectation argument as "superficial and conclusory," noting that "Bell and Shi are both aimed at 'casual video' stabilization for mobile devices… lacking the rigorous stabilization needed for more specialized applications."
2. Technical incompatibility / inoperability. GoPro argued the combination of an anticipatory look-ahead smoother with a reactive blur handler "undermine[s] Bell's look-ahead functionality."
3. No teaching or suggestion of the proactive parameter. GoPro argued "there is no clear teaching or suggestion in the references that would motivate a skilled artisan to combine Bell's smoothing technique with Shi's reactive, blur-handling method to arrive at a system incorporating the proactive low-light high-pass parameter."
4. Hindsight. GoPro expressly argued "Petitioner's reliance on the alleged combination of Bell and Shi reflects improper hindsight reasoning."
5. All dependents stand or fall together. GoPro argued Grounds 2–7 "fail for the same reason as Ground 1, as all dependent claims require one or more of the low-light high pass, weight-balance and stickiness parameters of the independent claims."

C. Why the Zhou + Cai case had a real fight. The disputed questions were narrower — whether Cai's window-of-frames operation is genuinely future-frame based (GoPro: "based on past and current frames only"); whether Zhou's Gaussian weighting is an adjustable stickiness parameter; and whether Zhou's blur-based stabilization-strength adjustment is a low-light high-pass parameter under the correct construction. The Board's institution decision resolved the Cai issue in petitioner's favor on a preliminary record; the FWD nonetheless went the other way.

D. The ITC overlay — a validity finding against claim 14. Separately from the PTAB:

• ALJ Final Initial Determination (July 10, 2025): claims 13 and 14 not shown to be infringed; claim 14 shown to be invalid; claim 13 not shown invalid; domestic industry satisfied. (Order No. 9 and Order No. 24 had previously narrowed the asserted '840 claims from 1–21 down to 13 and 14 only.)
• Commission Final Determination (Feb. 26, 2026): reversed the ID's finding that the accused cameras satisfy the "low-light high-pass parameter" and the additional limitations of claim 14 → no violation as to the '840 patent; the only violation found was under design patent D789,435 (limited exclusion order + cease-and-desist).

⚠️ Important unresolved point: I could not verify from the retrieved Federal Register text which statutory ground the ALJ used to invalidate claim 14 (obviousness under §103 vs. indefiniteness under §112), nor whether the Commission's reversal of the "additional limitations of claim 14" subsumed the invalidity finding. The low-light high-pass parameter is precisely the claim 14 limitation that both the ITC and the PTAB struggled with, which makes the ITC's invalidity finding on claim 14 highly relevant to — but not dispositive of — the §103 question, and a likely subject of the pending Federal Circuit appeals (26‑1392, 26‑1393, 26‑1752; ⚠️ docket→proceeding mapping unverified).

8. Bottom line

The strongest §103 combinations available against US 10,958,840 are, in descending order of strength:

1. Zhou (US 9,674,438) + Cai (US 10,158,802) — structurally the most complete mapping. Zhou alone supplies the housing/optical/sensor/IMU combination, the look-ahead-buffered smoothed trajectory, and the visible-frame-within-captured-frame punchout (i.e., elements 1[a]–1[i] except the intentional-motion temporal horizon). Cai supplies the bounded future-frame window that converts raw motion into an intentional/jitter classification. The motivation is the strongest of any pair here: both references are directed at the same sub-problem — telling intentional camera motion from shake — and Zhou already inspects "future frames."
2. Bell + Shi — strong on elements (Bell supplies nearly the entire apparatus claim including the crop-polygon punchout; Shi supplies the blur-tracking adjustment that reads on "low-light high-pass") but weakest on motivation to combine, because Shi's method is reactive and Bell's is anticipatory, which is exactly the "change the principle of operation / render inoperable" defense that defeated Ground 1.
3. Bell + Shi + Bell Paper (claims 6, 7, 21) and Zhou + Cai + Bell Paper (claims 13, 17) — the Bell Paper is the cleanest single-reference disclosure of a combined, weighted minimization of rotational velocity and acceleration, which is verbatim the '840 specification's own objective function. This is the most doctrinally comfortable obviousness case on the weight-balance parameter specifically.
4. + Karpenko (claims 4, 20) and + Sokeila (claim 12) — classic secondary-reference "known technique / finite design choice" grounds; the Sokeila/circular-buffer ground in particular is close to a pure KSR "obvious to try" case and is the least contestable.

But the analysis cannot conclude "obvious." The only adjudicated §103 determinations on this patent are:

• IPR2024‑01164 and IPR2024‑01165, FWDs of Nov. 25, 2025: no challenged claim held unpatentable (now on appeal); and
• ITC 337‑TA‑1400: claim 14 held invalid by the ALJ (ground unverified), claim 13 not invalid, and the Commission reversing the infringement findings on the low-light high-pass limitation → no §337 violation.

So the accurate statement of the law/fact as of today is: the prior art combinations are sufficient to have compelled institution in two separate IPRs and to have sustained an ITC invalidity finding on claim 14, but insufficient to invalidate any claim before the Board on a full record. The residual non-obviousness in this patent lives almost entirely in (i) the coined, narrowly-argued "low-light high-pass parameter", and (ii) the requirement that the look-ahead impair the output (the mandatory delay), rather than in any new smoothing mathematics.

Explicit uncertainties I will not paper over:

1. The patent/publication number of EX1006 "Bell" in IPR2024‑01164 is unverified.
2. Shi = US 10,462,370 is my inference, not a confirmed identification; I could not retrieve the petition's bibliographic line naming the number.
3. Sokeila: US 8,994,838 (per the IPR2024‑01165 ground listing) vs. US 9,215,377 (a Sokeila et al. patent that surfaced in search) — unresolved divergence.
4. IPR2024‑01164 Grounds 5–7 were not retrieved; I do not know which claims they covered.
5. The Board's actual FWD reasoning is not in my retrieved sources — Section 7's "why it failed" is reconstructed from briefs and hearing transcripts, not the opinion.
6. The statutory basis for the ITC's claim 14 invalidity finding is unverified.
7. CAFC dockets 26‑1392 / 26‑1393 / 26‑1752 are reported from Google Patents' litigation linkage; the mapping to the two IPR appeals and the ITC appeal is inferred, not confirmed.

Primary sources relied on: the authoritative Google Patents text for US 10,958,840; PTACTS petition/declaration/transcript excerpts at https://ptacts.uspto.gov/ptacts/public-informations/petitions/1556345/... (IPR2024‑01164) and .../1556346/... (IPR2024‑01165); institution-decision and case-status records at https://ipverse.greyb.com/ptab-web/cases/case-details/IPR2024-01164 and .../IPR2024-01165; the Ex Parte AI-lab petition summaries at https://ai-lab.exparte.com/case/ptab/IPR2024-01165/...; the ITC final determination at https://www.govinfo.gov/content/pkg/FR-2026-03-03/pdf/2026-04156.pdf; the ITC Final ID notice at https://www.usitc.gov/press_room/documents/notice_of_issuance_final_idrd_1400.pdf; and the independent Chinese-language report of the Nov. 25, 2025 FWDs at https://news.qq.com/rain/a/20251126A06V4000.