ZEETAD: Adapting Pretrained Vision-Language Model for Zero-Shot End-to-End Temporal Action Detection.
Thinh PhanKhoa VoDuy LeGianfranco DorettoDonald A. AdjerohNgan LePublished in: CoRR (2023)
Keyphrases
- end to end
- language model
- action detection
- language modeling
- n gram
- probabilistic model
- information retrieval
- action recognition
- query expansion
- retrieval model
- temporal information
- context sensitive
- atomic actions
- test collection
- mixture model
- temporal reasoning
- spatio temporal
- computer vision
- temporal relations
- action classification
- pattern search
- human actions
- graphical models
- data streams
- human activity recognition