Adversarial Augmentation Training Makes Action Recognition Models More Robust to Realistic Video Distribution Shifts.
Kiyoon KimShreyank N. GowdaPanagiotis EustratiadisAntreas AntoniouRobert B. FisherPublished in: CoRR (2024)
Keyphrases
- action recognition
- human actions
- spatial temporal
- action classification
- action detection
- video dataset
- view invariant
- recognizing human actions
- bag of words
- recognition of human actions
- computer vision
- motion features
- human activities
- human detection
- mid level
- multimedia
- activity recognition
- space time
- video data
- static images
- structured prediction
- video sequences
- space time interest points
- video analysis
- body parts
- video frames
- human body