TrouSPI-Net: Spatio-temporal attention on parallel atrous convolutions and U-GRUs for skeletal pedestrian crossing prediction.
Joseph GesnouinSteve PechbertiBogdan StanciulescuFabien MoutardePublished in: CoRR (2021)
Keyphrases
- spatio temporal
- prediction accuracy
- spatial and temporal
- prediction model
- human motion
- spatial temporal
- parallel implementation
- parallel processing
- spatio temporal patterns
- real time
- crowded scenes
- prediction error
- space time
- object detection
- multiscale
- scale space
- motion estimation
- human actions
- moving objects
- object recognition
- video sequences
- massively parallel
- parallel computation
- prediction algorithm
- image sequences
- three dimensional