1 video input, 1 infer task, and 1 output.
1 video input, 1 infer task, and 2 outputs.
1 video input and then split into 2 branches for different infer tasks, then 2 total outputs.
2 video input and merge into 1 branch automatically for 1 infer task, then resume to 2 branches for outputs again.
multi pipe exist separately and each pipe is 1-1-1 (can be any structure like 1-1-N, 1-N-N)
ocr based on paddle (install paddle_inference first!), 1 video input and 2 outputs (screen, rtmp)
show how src nodes and des nodes work.
3 (file, rtsp, udp) input and merge into 1 infer task, then resume to 3 branches for outputs (screen, rtmp, fake)
vehicle and plate detector based on tensorrt (install tensorrt first!), 1 video input and 3 outputs (screen, file, rtmp)
show how vp_logger works.
tracking for multi faces.
tracking for multi vehicles.
show how to interact with pipe, such as start/stop channel by calling api.
show how vp_record_node works.
show how message broker nodes work.
show image segmentation by mask-rcnn.
show pose estimation by openpose network.
show semantic segmentation by enet network.
show multi infer node work together.
show save/push image to local file or remote via udp.
show read/receive image from local file or remote via udp.
show push video stream via rtsp, no rtsp server needed, you can visit it directly.
count for vehicle based on tracking, the simplest one of behaviour analysis.
vehicle plate detect and recognize on the whole frame (no need to detect vechile first)
