Introduction
The purpose of this project is trying to do multiple objects tracking using YOLOv3 with DeepSORT, and make the people tracking more robust. Thanks for these projects, this project now can support YOLOv3.
https://github.com/nwojke/deep_sort
https://github.com/qqwweee/keras-yolo3
https://github.com/Qidian213/deep_sort_yolov3
Some excellent related work
Quick Start
-
Download YOLOv3 or tiny_yolov3 weights and cfg from YOLO website, and put it into model_data folder.
-
Put the video into the test_video folder, and modify the video_file name in
demo.py
. -
Run YOLO_DEEP_SORT with command :
python demo.py
Training the model
To train the deep association metric model on your datasets you can reference to cosine_metric_learning approach which is provided as a separate repository.
Result Video
- Result video on MOT16-13
- Result video on Taiwan street