TrouSPI-Net: Spatio-temporal attention on parallel atrous convolutions and U-GRUs for skeletal pedestrian crossing prediction.
Joseph GesnouinSteve PechbertiBogdan StanciulcscuFabien MoutardePublished in: FG (2021)
Keyphrases
- multimedia
- spatio temporal
- prediction accuracy
- spatial and temporal
- image sequences
- spatio temporal patterns
- object detection
- moving objects
- spatio temporal data
- parallel processing
- neural network
- prediction algorithm
- pedestrian detection
- human motion
- spatio temporal databases
- focus of attention
- parallel implementation
- prediction model
- visual attention
- spatial temporal
- prediction error
- parallel computation
- space time
- image processing