Strike (with) a Pose: Neural Networks Are Easily Fooled by Strange Poses of Familiar Objects
CVPR, Volume abs/1811.11553, 2019, Pages 4845-4854.
Despite excellent performance on stationary test sets, deep neural networks (DNNs) can fail to generalize to out-of-distribution (OoD) inputs, including natural, non-adversarial ones, which are common in real-world settings. In this paper, we present a framework for discovering DNN failures that harnesses 3D renderers and 3D models. That ...More
PPT (Upload PPT)