Offical Implementation of Conference Paper Published in 8th International Conference on Signal and Image Processing (SIGPRO 2022). Toronto, Canada

Lead Authors

Saira Gillani (https://www.linkedin.com/in/saira-gillani/)
Muhammad Rizwan Munawar (https://pk.linkedin.com/in/muhammadrizwanmunawar)

Conference Paper Link: Yolov5, Yolo-x, Yolo-r, Yolov7 Performance Comparison: A Survey

Object Detection Algorithms YOLO Series [Suppport for Windows & Linux]

FPS-Comparision-YOLO-Series

Leading YOLO Series Algorithms FPS Comparision with same specifications.
The comparision include, YOLOv7, YOLOv5, YOLOX, YOLO-R

Recommended Envirnoment

Operating System : Ubuntu 18/Ubuntu 20/Windows 10
python 3.6/3.7/3.8

Results

Start Guide

For Ubuntu

git clone https://github.com/RizwanMunawar/YOLO-RX57-FPS-Comparision.git
cd YOLO-RX57-FPS-Comparision
pip install Cython opencv-python
sudo apt update
sudo apt install build-essential
sudo apt-get install manpages-dev
pip install "git+https://github.com/philferriere/cocoapi.git#egg=pycocotools&subdirectory=PythonAPI"

Now you will need to download weights of YOLOR, YOLOv7 and YOLOX and move downloaded weights into respective folders using mentioned commands below.

cd YOLORResearch/yolor
wget https://github.com/RizwanMunawar/YOLO-RX57-FPS-Comparision/releases/download/v0.1.4-release/yolor_p6.pt
cd ..
cd ..
cd YOLOXResearch/YOLOX
wget https://github.com/RizwanMunawar/YOLO-RX57-FPS-Comparision/releases/download/v0.1.4-release/yolox_s.pth
cd ..
cd ..
wget YOLOv7Research/yolov7
wget https://github.com/RizwanMunawar/YOLO-RX57-FPS-Comparision/releases/download/v0.1.4-release/yolov7.pt
cd ..
cd ..

Now you can run code with mentioned command below.

python3 comparision.py --source "2.mp4" --device 0

or

python3 comparision.py --source "2.mp4" --device cpu

In above command, source will be the path of your video/image on, which you want to calculate FPS.

The above command will store the output of all detectors in a new folder named results. The results folder will contain output videos for every object detector, FPS will be shown on every video.

If you want to use your custom weights, you can use mentioned command below.

python3 comparision.py --source "2.mp4" --device 0 \
                       --yolov5weights yolov5s.pt --yolorweights yolor_p6.pt \
                       --yoloxweights yolox_m.pt --yolov7weights yolov7.pt

For Windows 10

git clone https://github.com/RizwanMunawar/YOLO-RX57-FPS-Comparision.git
cd YOLO-RX57-FPS-Comparision
pip install pycocotools-windows

Now you will need to download weights of YOLOR, YOLOv7 and YOLOX and move downloaded weights into respective folders. Mentioned steps below to complete this.

Download yolor_p6.pt file and move the downloaded file in YOLORResearch/yolor folder
Download yolov7.pt file and move the downloaded file in YOLOv7Research/yolov7 folder
Download yolox_s.pth file and move the downloaded file in YOLOXResearch/YOLOX folder

Now you can run code with mentioned command below.

python3 comparision.py --source "2.mp4" --device 0

or

python3 comparision.py --source "2.mp4" --device cpu

In above command, source will be the path of your video/image on which you want to calculate FPS.

The above command will store the output of all detectors in a new folder named results. The results folder will contain output videos for every object detector, FPS will be shown on every video.

If you want to use your custom weights, you can use mentioned command below.

python3 comparision.py --source "2.mp4" --device 0 \
                      --yolov5weights yolov5s.pt --yolorweights yolor_p6.pt \
                      --yoloxweights yolox_m.pt --yolov7weights yolov7.pt