Basic video object detection using OpenCV and a tiny yolo network.

• Numpy
• OpenCV

pip install numpy opencv-python

python3 app.py -i ./video_file.mp4 -s 
  or
python3 app.py -i ./video_file.mp4 -s -c ./cfg/tiny-yolov4.cfg -w ./cfg/yolov4-tiny.weights -o video_det 

python3 app.py -i ./video_file.mp4 -s -l
  or
python3 app.py -i ./video_file.mp4 -s -c ./cfg/tiny-yolov4.cfg -w ./cfg/yolov4-tiny.weights \ 
   -o video_det -l -lf labels_output -sl 10

Options:

• -i or --input-uri : input file or stream eg. "video_file.mp4" or "192.1.20.4/live"
• -s or --show : display opencv window with detections/tracks
• -w or --weights : yolo weights file path eg. ./cfg/tiny-yolov4.cfg
• -c or --config: yolo configs file path eg. ./cfg/yolov4-tiny.weights
• -o or --output-uri: video output filename, saves the file to ./output folder
• -l or --label: perform pseudo-labeling on detections
• -lf or --label-folder: folder to create for saving the pseudo-labels (images and txt)
• -sl or --skip-label: frames to skip when performing pseudo-labeling

Class names can be found in './cfg/coco.names'