I implement the method Seq-NMS mentioned in paper: SEQ-NMS FOR VIDEO OBJECT DETECTION based on the Faster-RCNN. As for the detection in a single frame, I use the demo program and pre-trained network ...