Multi Tasks(Detect, Track, Dense, Count) in One Frame-Work

A multi task frame-work based on Yolov5+Deepsort, contains:

• detect task
• track task
• dense estimate task
• object counting task

Demo

dense estimate demo:

object counting demo:

tracking demo(with velocity visulization):

Requirements

Installation

1.clone this repository

git clone https://gitlab.10010sh.cn/ai/yolov5_deepsort
cd yolov5_deepsort

2.download yolov5 weights

cd pytorch_yolov5/weights

download weights file(yolov5l.pt) from yolov5 V2.0 (at the bottom) to this folder.

cd ../../

3.download deepsort weights

cd deepsort/deepsort/deep/checkpoint

download weights file(ckpt.t7) from deepsort ckpt to this folder.

cd ../../../../

Usage

python main.py --task detect --input {path to images or video or camera} --output {path to result save folder}
                      track
                      dense
                      count

more detail parameters can seen in main.py

References

Thanks for the great work from [yolov5] and [deepsort].