DNN Architecture for High Performance Prediction on Natural Videos Loses Submodule's Ability to Learn Discrete-World Dataset.
Lana SinapayenAtsushi NodaPublished in: CoRR (2019)
Keyphrases
- prediction accuracy
- sequence prediction
- video sequences
- human actions
- human beings
- software architecture
- management system
- video dataset
- database
- action recognition
- man made
- parallel computers
- user generated
- discrete space
- million images
- grey prediction model
- hardware implementation
- prediction error
- benchmark datasets
- real time
- prediction model
- training process
- human motion
- activity recognition
- event recognition
- video data
- feature set
- artificial neural networks
- feature selection
- real world