Prior art

To identify the most relevant prior art for US patent 11195058, I will examine the patent's citations. However, I am unable to directly access and parse the full list of cited prior art from the USPTO database for US11195058 within this environment. The provided patent text does not explicitly list the "Prior Art Citations" section in a parseable format, and direct access to live USPTO search results to extract this specific information is not available to me.

Therefore, I cannot provide a detailed analysis of each prior art reference, including full citation, publication/filing date, brief description, and which claim(s) it potentially anticipates under 35 U.S.C. § 102.

However, based on the Background section of US11195058, it is generally understood that existing "risk exposure databases" compile various building properties and characteristics relevant to insurance, some of which can be measured using visual imagery. The patent also acknowledges the use of deep learning, citing "Network In Network" by M. Lin et al. (published in the International Conference on Learning Representations, 2014) as a model used for deep learning. This indicates that general concepts of deep learning and its application in visual recognition for feature characterization were known prior to this patent's filing.

The core innovation of US11195058, as highlighted in the litigation summary (Aon Re, Inc. v. Zesty.AI, Inc., No. 1:25-cv-00201-JFM (D. Del. July 15, 2025)), lies in its "distinctive structure of the process involving two separate machine-learning classifiers" for analyzing images to determine property characteristic classifications and condition classifications, which then informs risk estimates. This suggests that prior art that only uses a single classifier or lacks this specific two-classifier architecture for both characteristic identification and condition assessment would be less anticipatory of the patent's claims, particularly Claim 1 which covers this method.