RNN-Guard: Certified Robustness Against Multi-frame Attacks for Recurrent Neural Networks.
Yunruo ZhangTianyu DuShouling JiPeng TangShanqing GuoPublished in: CoRR (2023)
Keyphrases
- recurrent neural networks
- multi frame
- motion estimation
- point correspondences
- image sequences
- neural network
- multiple frames
- computationally efficient
- long short term memory
- super resolution
- motion analysis
- temporal coherence
- recurrent networks
- feed forward
- single frame
- optical flow
- artificial neural networks
- echo state networks
- optic flow
- neural model
- reservoir computing
- filtering algorithm
- cascade correlation
- video frames
- high quality
- artificial intelligence
- signal to noise ratio
- nonlinear dynamic systems
- genetic algorithm