Strike (With) a Pose: Neural Networks Are Easily Fooled by Strange Poses of Familiar Objects.
Michael A. AlcornQi LiZhitao GongChengfei WangLong MaiWei-Shinn KuAnh NguyenPublished in: CVPR (2019)
Keyphrases
- neural network
- d objects
- human pose
- object geometry
- pose invariant face recognition
- pose estimation
- pattern recognition
- moving objects
- genetic algorithm
- neural network model
- object representation
- object recognition
- back propagation
- object model
- focus of attention
- articulated objects
- cluttered background
- human body
- partial occlusion
- neural nets
- training process
- viewpoint
- kinematic constraints
- computer vision