本代码库实现视频浓缩功能，将视频不同时间段出现的人物浓缩在较短的时间内。
This code library implements video condensation function, which condenses characters that appear in different time periods of the video in a relatively short amount of time.
- You need to first obtain "yolov4. weights" and place it in the "./data" folder.
Then run:
python save_model.py The "./checkpoints" folder will be generated.
- run
python main.py
All intermediate files and final condensed videos will be generated in the "./outputs" folder.
Original video:
demo：（The file is located in the output folder）
<video src="outputs/demo_with_boxes.mp4"></video>
The first frame of the original video：
The first frame of the new video：
It is evident that the first frame of the new video contains the following additional targets: person-40 (appearing in frame 227 of the original video), person-43 (appearing in frame 293 of the original video), person-16 (appearing in frame 8 of the original video), person-35 (appearing in frame 207 of the original video), and person-15 (appearing in frame 6 of the original video) with only half of its legs in the upper right corner.