mikel-brostrom/yolov3_deepsort_pytorch

Real-time multi-person tracker using YOLO v3 and deep sort

Yolov3 + Deep Sort with PyTorch

Introduction

This repository contains a moded version of PyTorch YOLOv3 (https://github.com/ultralytics/yolov3). It filters out every detection that is not a person. The detections of persons are then passed to a Deep Sort algorithm (https://github.com/ZQPei/deep_sort_pytorch) which tracks the persons. The reason behind the fact that it just tracks persons is that the deep association metric is trained on a person ONLY datatset.

Description

The implementation is based on two papers:

• Simple Online and Realtime Tracking with a Deep Association Metric https://arxiv.org/abs/1703.07402
• YOLOv3: An Incremental Improvement https://arxiv.org/abs/1804.02767

Requirements

Python 3.7 or later with all of the pip install -U -r requirements.txt packages including:

• torch >= 1.3
• opencv-python
• Pillow

All dependencies are included in the associated docker images. Docker requirements are:

• nvidia-docker
• Nvidia Driver Version >= 440.44

Before you run the tracker

Github block pushes of files larger than 100 MB (https://help.github.com/en/github/managing-large-files/conditions-for-large-files). Hence the yolo weights needs to be stored somewhere else. When you run tracker.py you will get an exceptions telling you that the yolov3 weight are missing and a link to download them from. Place the downlaoded .pt file under yolov3/weights/. The weights for deep sort are already in this repo. They can be found under deep_sort/deep/checkpoint/.

Tracking

track.py runs tracking on any video source:

python3 track.py --source ...

• Video: --source file.mp4
• Webcam: --source 0
• RTSP stream: --source rtsp://170.93.143.139/rtplive/470011e600ef003a004ee33696235daa
• HTTP stream: --source http://wmccpinetop.axiscam.net/mjpg/video.mjpg

Cite

If you find this project useful in your research, please consider cite:

@misc{yolov3-deepsort,
    title={Real-time multi-camera multi-object tracker using YOLOv3 and DeepSORT},
    author={Mikel Broström},
    howpublished = {\url{https://github.com/mikel-brostrom/Yolov3_DeepSort_Pytorch}},
    year={2019}
}

Other information

For more detailed information about the algorithms and their corresponding lisences used in this project access their official github implementations.

yolov3_deepsort_pytorch's Issues

No module named 'models'

I am having trouble running the file and hope someone can help me.
When I run the track.py, I get the following:

F:\OOHAN\ZhuXY-Projects\MOT\DeepSort-mikel-brostrom\Yolov3_DeepSort_Pytorch>python track.py --source test.mp4
Loading weights from deep_sort/deep/checkpoint/ckpt.t7... Done!
return _load(opened_zipfile, map_location, pickle_module, **pickle_load_args)
File "C:\python\anzhuang\envs\OOHANpytorch\lib\site-packages\torch\serialization.py", line 882, in _load
result = unpickler.load()
File "C:\python\anzhuang\envs\OOHANpytorch\lib\site-packages\torch\serialization.py", line 875, in find_class
return super().find_class(mod_name, name)
ModuleNotFoundError: No module named 'models'

Optimizing tracking script, a small change that can push tracking from lower fps to atleast 30 fps

During inference there's a small mistake you've done or maybe you didn't notice it.
when you check the detected output of yolo that is 'det'. you update the tracker using a list. now the thing you've missed is you push xywh list again and again into the tracker because of the iteration. all you have to do is do this push the entire list once. I've attached an image so you'll understand. just move the indent of deepsort.update and the next if loop which consist of drawing bounding box. this will make updating tracker and drawing bounding box n (number of objects detected in frame) times faster.

模型链接失效

你好，
Downloading https://pjreddie.com/media/files/yolov3-spp-ultralytics.pt

链接失效，能提供百度云链接的.pt文件吗，非常感谢

Exception: yolov3/weights/yolov3-spp-ultralytics.pt missing, try downloading from https://drive.google.com/open?id=1LezFG5g3BCW6iYaV89B2i64cqEUZD7e0

🐛 Bug

the link is 404 not found

About training datasets

Hi there, really appreciate ur sharing. Could u tell me what datasets u used when training the yolo in this person-only mode?

ide

🐛 Bug

A clear and concise description of what the bug is.

To Reproduce

Steps to reproduce the behavior:
1.
2.
3.

Expected behavior

A clear and concise description of what you expected to happen.

Environment

If applicable, add screenshots to help explain your problem.

Desktop (please complete the following information):

• OS: [e.g. iOS]
• Version [e.g. 22]

Smartphone (please complete the following information):

• Device: [e.g. iPhoneXS]
• OS: [e.g. iOS8.1]
• Version [e.g. 22]

Additional context

Add any other context about the problem here.

Link to yolov3 pytorch weights file cannot be accessed

Hi, I was trying to use the trained weights on my own model and was unable to access the file from google drive.

I was wondering if this file could be made public again?

Thank you!

How to solve this error? RuntimeError: Could not run 'torchvision::nms' with arguments from the 'CUDA' backend

RuntimeError: Could not run 'torchvision::nms' with arguments from the 'CUDA' backend. 'torchvision::nms' is only available for these backends: [CPU, BackendSelect, Named, AutogradOther, AutogradCPU, AutogradCUDA, AutogradXLA, Tracer, Autocast, Batched, VmapMode].

Actually, I didn't get this error and couldn't search anything about this on google.
Is there anybody who knows the way to handle this?

Info:
I installed the CUDA v11.1 from https://developer.nvidia.com/cuda-downloads
torch version: 1.7.0
torchvision version: 0.8.0

How to use device GPU

Hello,
I wanted to use the tracker with the GPU that I have on my machine, I dont want to use the docker file. I tried providing the --device argument but it doent work. I have a single GPU machine with Nvidia GTX 1060.

May I know how to create confusion matrix ?

🚀 Feature

Motivation

Pitch

Alternatives

Additional context