honglinchu / yolov3-tiny-deepsort-pytorch-cardetection

yolov3_tiny(add SE model)(pytorch 1cls for car),deep_sort(pytorch),mx150 GPU, 14 avg_fps

Deep Sort and yolov3_tiny for car detection with PyTorch

nearly realtime:avg_fps=14 for mx150(2G GPU)

Latest Update(2020-3-9)

Weights files and demo.mp4 at here:(also i put this video to bilibili,after passing the audit,i will put 链接) [BaiduDisk] 链接：https://pan.baidu.com/s/1czw0dz56ZjfgrxjBFaUSdg 提取码：odbu Yolov3_tiny_1cls_se detection results at here: 链接：https://www.bilibili.com/video/av94365366/

Dependencies

• python 3
• numpy
• scipy
• opencv-python
• sklearn
• torch ==1.1.0
• torchvision ==0.3.0
• pillow
• vizer
• edict

Quick Start

0. Check all dependencies installed

1. Download best.pt(yolo weights) parameters

look up

3. Download deepsort parameters ckpt.t7

look up and put here:
cd deep_sort/deep/checkpoint

Notice: If compiling failed, the simplist way is to **Upgrade your pytorch >= 1.1 and torchvision >= 0.3" and you can avoid the troublesome compiling problems which are most likely caused by either gcc version too low or libraries missing.

4. Run my_deep_sort.py

my is windows
i do not know linux or ubuntu
my programming ability is very weak
comments are not very good
forgive me

Notice: Maybe some mistakes, you can discuss in issues, help me to modify. It is just a small work, you can improve based on my code.

References

• paper: Simple Online and Realtime Tracking with a Deep Association Metric

• code: nwojke/deep_sort

• paper: YOLOv3

• code: ultralytics/yolov3

• code:ZQPei/deep_sort_pytorch