Yolov5 + Deep Sort with PyTorch

Introduction

This repository contains a moded version of PyTorch YOLOv5 (https://github.com/ultralytics/yolov5).

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
• YOLOv4: Optimal Speed and Accuracy of Object Detection https://arxiv.org/pdf/2004.10934.pdf

Requirements

Python 3.8 or later with all requirements.txt dependencies installed, including torch>=1.6. To install run:

pip install -U -r requirements.txt

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

• nvidia-docker
• Nvidia Driver Version >= 440.44

Before you run the tracker

1. Clone the repository recursively:

git clone --recurse-submodules https://github.com/mikel-brostrom/Yolov5_DeepSort_Pytorch.git

If you already cloned and forgot to use --recurse-submodules you can run git submodule update --init

2. Github block pushes of files larger than 100 MB (https://help.github.com/en/github/managing-large-files/conditions-for-large-files). Hence you need to download two different weights: the ones for yolo and the ones for deep sort

• download the yolov5 weight from the latest realease https://github.com/ultralytics/yolov5/releases. Place the downlaoded .pt file under yolov5/weights/
• download the deep sort weights from https://drive.google.com/drive/folders/1xhG0kRH1EX5B9_Iz8gQJb7UNnn_riXi6. Place ckpt.t7 file underdeep_sort/deep/checkpoint/

Tracking On Screen

The original tracking can be run on most video formats, but I created a new one which can be applied to many Industrial scenarios, that is analyze a live streaming video displayed on your screen.

python tracking_on_screen(trajectory).py

Parameter Interpretation

--weights: yolov5/weights/yolov5s.pt
Path of the model you want to deploy.

--config-deepsort: deep_sort_pytorch/configs/deep_sort.yaml
A yaml file path that contains all the deepsort setup.

--Record:
By default, the tracking and detecting results would not be saved into a text file as a format of continuous numpy arrays.

--record-path: inference/output
Path for the result text file.

--record-name: tracking_record.txt

--device:
Cuda device, gpu 0 or 1, 2, 3. Or if you are running without cuda, you should set it to cpu.

--conf-thres:
Object confidence threshold.

--iou-thres:
IOU threshold for NMS.

--classes: [0]
Choose the corresponding class id that you want to track in you model.

The following 4 parameters defines the region in your computer monitor.
--region-x1: 0
--region-y1: 175
--region-x2: 928
--region-y2: 687

The following 2 parameters defines how you want to resize the images for processing. For the constriction of resize-width and resize-height, you should also need to check the 11-23-2020 report for instruction.
--resize-width: 928
--resize-height: 512

Other information

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