Artificial Intelligence Industry Association Inc v. Geisel Software Inc

I. Executive Summary and Procedural Information

• Parties & Counsel:
  • Plaintiff: Artificial Intelligence Industry Association, Inc. (Florida)
  • Defendant: Geisel Software, Inc. (Delaware)
  • Plaintiff’s Counsel: Michael G. Newell
• Case Identification: 1:25-cv-07387, S.D.N.Y., 12/24/2025
• Venue Allegations: Plaintiff alleges venue is proper because Defendant has committed acts of infringement in the district by offering for sale and selling its accused software products to customers with operations in New York, such as the iRobot Corporation.
• Core Dispute: Plaintiff alleges that Defendant’s Symage AI platform for generating synthetic data infringes a patent related to systems and methods for creating labeled image data for machine learning.
• Technical Context: The technology at issue involves the automated generation of synthetic, photorealistic images and associated data (e.g., depth maps) used to train computer vision models for tasks like object detection and image segmentation.
• Key Procedural History: The complaint alleges that prior to filing suit, Plaintiff sent Defendant a formal demand letter identifying the asserted patent and its alleged infringement, which serves as the basis for the willfulness allegation.

Case Timeline

II. Technology and Patent(s)-in-Suit Analysis

U.S. Patent No. 11,257,272 - "Generating Synthetic Image Data for Machine Learning"

• Patent Identification: U.S. Patent No. 11,257,272, "Generating Synthetic Image Data for Machine Learning," issued February 22, 2022.

The Invention Explained

• Problem Addressed: The patent describes a technical problem in the field of computer vision: training accurate machine learning models requires vast amounts of specialized image data (e.g., stereo image pairs with corresponding depth information), but manually capturing such data is expensive, time-consuming, and often results in low-quality or limited datasets (’272 Patent, col. 2:5-40).
• The Patented Solution: The invention provides a system and method to automatically generate large, diverse, and richly annotated synthetic image datasets. The process involves programmatically constructing a virtual 3D scene from a database of components (e.g., background images, 3D models, textures) and then using a "virtual camera" to capture various perspectives. These virtual cameras are defined by camera settings files that can simulate the performance and specific parameters of real-world cameras, thereby creating realistic training data (’272 Patent, Abstract; Fig. 11).
• Technical Importance: This automated approach addresses the data bottleneck in machine learning development by enabling the rapid, scalable, and cost-effective creation of high-quality, customized training datasets for computer vision tasks (’272 Patent, col. 7:1-19).

Key Claims at a Glance

• The complaint asserts independent method claim 17 (’272 Patent, col. 51:17-52:32; Compl. ¶16).
• The essential elements of Claim 17 include:
  • Receiving databases of background images, 3D models, textures, scene metadata, and camera setting files.
  • Constructing a first synthetic image scene with a specified image scene class using a computer graphics engine.
  • Placing a virtual camera to capture a series of camera views.
  • Rendering projection coordinates as synthetic images for each camera view.
  • Constructing a second synthetic image scene with the same scene class and capturing views with the same camera positions.
• The complaint alleges infringement of "one or more claims," including at least Claim 17 (Compl. ¶19).

III. The Accused Instrumentality

Product Identification

• The accused products include Defendant's "Symage AI" platform and associated "Synthetic Data Generation Platforms, APIs, and tools for Dataset Creation" (Compl. ¶1).

Functionality and Market Context

• The complaint alleges the accused products are a "synthetic image data generator" that creates "physics-based synthetic image data that's flawlessly labeled and photorealistic" for training machine learning models (Compl. ¶21). The platform is alleged to use "3D simulation and virtual camera configurations" to generate customizable scenes with 3D models and varied camera perspectives, along with additional data channels like depth maps (Compl. ¶1, ¶20). Plaintiff asserts that it and Defendant compete directly in the market for synthetic data generation solutions for machine learning model training (Compl. ¶27).

IV. Analysis of Infringement Allegations

U.S. Patent No. 11,257,272 Infringement Allegations

No probative visual evidence provided in complaint.

• Identified Points of Contention:
  • Scope Questions: A central question may be whether the accused platform's user-selectable categories like "cityscape, landscape" (Compl. ¶20) meet the claim limitation of "a specified image scene class," which the patent links to instructions in a "scene metadata file" ('272 Patent, col. 51:8-14). The dispute could turn on whether "scene class" requires a formal, programmatic definition or if a user-selected theme suffices.
  • Technical Questions: The complaint alleges the platform generates additional scenes with the "same scene class and camera positions" (Compl. ¶20). A technical question is what evidence will show that the accused process for generating subsequent scenes adheres to the specific constraints of using the exact same camera positions and class as the first scene, as required by the final two steps of Claim 17.

V. Key Claim Terms for Construction

• The Term: "image scene class"
• Context and Importance: This term appears in the core steps of constructing the first and second synthetic scenes. Its construction is critical because infringement may depend on whether the Defendant's method of categorizing scenes (e.g., by "cityscape" or "landscape" types) falls within the patent's definition. Practitioners may focus on this term because it appears to be a key organizing principle for the claimed method.
• Intrinsic Evidence for Interpretation:
  • Evidence for a Broader Interpretation: The patent specification provides a list of exemplary image scene classes, including "landscape scene, cityscape scene, selfie scene, wildlife scene, event scene, action scene, street view scene, and driver point of view scene" ('272 Patent, col. 52:30-32). This list, which includes general categories, could support an interpretation that covers the types of scenes allegedly offered by the Defendant.
  • Evidence for a Narrower Interpretation: Claim 17 recites constructing a scene with a class "specified in a first scene metadata file" ('272 Patent, col. 51:10-11). The specification further describes scene metadata files as including "scene assembly algorithms" and "composition information" ('272 Patent, col. 29:55-62). This language could support a narrower construction requiring the "scene class" to be a formal, machine-readable parameter defined within a metadata file that dictates scene construction, rather than just a high-level descriptive category.

VI. Other Allegations

• Indirect Infringement: The complaint alleges inducement under 35 U.S.C. § 271(b) based on Defendant providing "tutorials, user guides, and other implementation resources on its website" that allegedly instruct customers on how to use the software in an infringing manner (Compl. ¶24). It also alleges contributory infringement under § 271(c), asserting that Defendant's software modules are "specifically designed and marketed for synthetic image generation" and have "no substantial non-infringing use" (Compl. ¶26).
• Willful Infringement: The complaint alleges willful infringement based on Defendant's continued infringing conduct after receiving actual notice of the ’272 Patent via a formal demand letter sent by Plaintiff prior to the lawsuit (Compl. ¶2, ¶31).

VII. Analyst’s Conclusion: Key Questions for the Case

• A core issue will be one of definitional scope: can the term "image scene class," which the patent ties to specifications within a "metadata file," be construed to cover the user-selectable scene categories like "cityscape" and "landscape" allegedly offered by the accused platform?
• A key evidentiary question will be one of procedural fidelity: what evidence will be required to demonstrate that the accused product, when generating subsequent images, constructs a "second synthetic image scene" and captures views from the "same camera positions" as the first, thereby satisfying the final, specific, and sequential limitations of asserted Claim 17?