Non-Driving Activities (NDAs) recognition using Two-stream ConvNets and FlowNet2.0
Identification of non-driving activities based on driver’s gesture using. Development of an artificial-intelligence-based solution to identify non-driving activities in Level 3 automated vehicle. Focusing on using Two-Stream Convolutional Neural Network, spatial and temporal data to enhance the performance of non-driving activities recognition. In order to obtain the driver’s movement information, FlowNet 2.0 was implemented to compute the optical flow of the driver.