3DB: A Framework for Debugging Computer Vision Models
This provides a tool for researchers and practitioners to improve computer vision model robustness, though it is incremental as it builds on prior robustness analyses.
The authors tackled the problem of testing and debugging computer vision models by introducing 3DB, a framework using photorealistic simulation, which allows users to discover vulnerabilities and gain insights into model decisions, with findings that transfer to the physical world.
We introduce 3DB: an extendable, unified framework for testing and debugging vision models using photorealistic simulation. We demonstrate, through a wide range of use cases, that 3DB allows users to discover vulnerabilities in computer vision systems and gain insights into how models make decisions. 3DB captures and generalizes many robustness analyses from prior work, and enables one to study their interplay. Finally, we find that the insights generated by the system transfer to the physical world. We are releasing 3DB as a library (https://github.com/3db/3db) alongside a set of example analyses, guides, and documentation: https://3db.github.io/3db/ .