YOLOv7 Object Tracking 🚀

How to Run the Code

1. Clone the repository:

   git clone https://github.com/RizwanMunawar/yolov7-object-tracking.git

2. Navigate to the cloned folder:

   cd yolov7-object-tracking

3. Create a virtual environment (Recommended to avoid conflicts):

   For Anaconda:

   conda create -n yolov7objtracking python=3.10
   conda activate yolov7objtracking

   For Linux:

   python3 -m venv yolov7objtracking
   source yolov7objtracking/bin/activate

   For Windows:

   python3 -m venv yolov7objtracking
   cd yolov7objtracking/Scripts
   activate

4. Update pip and install dependencies:

   pip install --upgrade pip
   pip install -r requirements.txt

5. Run the script:

   Select the appropriate command based on your requirements. Pretrained yolov7 weights will be downloaded automatically if needed.

   • Detection only:

     python detect.py --weights yolov7.pt
     
     # If you want to use your own videos.
     python detect.py --weights yolov7.pt --source "your video.mp4"

   • Object tracking:

     python detect_and_track.py --weights yolov7.pt
     
     # If you want to use your own videos.
     python detect_and_track.py --weights yolov7.pt --source "your video.mp4"

   • Webcam:

     python detect_and_track.py --weights yolov7.pt --source 0

   • External Camera:

     python detect_and_track.py --weights yolov7.pt --source 1

   • IP Camera Stream:

     python detect_and_track.py --source "your IP Camera Stream URL" --device 0

   • Specific class tracking (e.g., person):

     python detect_and_track.py --weights yolov7.pt --source "your video.mp4" --classes 0

   • Colored tracks:

     python detect_and_track.py --weights yolov7.pt --source "your video.mp4" --colored-trk

   • Save track centroids, IDs, and bounding box coordinates:

     python detect_and_track.py --weights yolov7.pt --source "your video.mp4" --save-txt --save-bbox-dim

6. Output files will be saved in working-dir/runs/detect/obj-tracking with the original filename.

Arguments details 🚀

Argument	Type	Default	Description
--weights	str	yolov7.pt	Path(s) to model weights (.pt file).
--download	flag	False	Download model weights automatically.
--no-download	flag	False	Do not download model weights if they already exist.
--source	str	None	Source for inference (file, folder, or 0 for webcam).
--img-size	int	640	Inference image size in pixels.
--conf-thres	float	0.25	Object confidence threshold.
--iou-thres	float	0.45	Intersection over Union (IoU) threshold for NMS.
--device	str	''	CUDA device (e.g., 0 or 0,1,2,3) or cpu.
--view-img	flag	False	Display results during inference.
--save-txt	flag	False	Save results to .txt files.
--save-conf	flag	False	Save confidence scores in .txt labels.
--nosave	flag	False	Do not save images or videos.
--classes	list[int]	None	Filter results by class (e.g., --classes 0 or --classes 0 2 3).
--agnostic-nms	flag	False	Use class-agnostic Non-Maximum Suppression (NMS).
--augment	flag	False	Enable augmented inference.
--update	flag	False	Update all models.
--project	str	runs/detect	Directory to save results (project/name).
--name	str	runs/detect/object_tracking	Name of the results folder inside the project directory.
--exist-ok	flag	False	Allow existing project/name without incrementing.
--no-trace	flag	False	Do not trace the model during export.
--colored-trk	flag	False	Assign a unique color to each track for visualization.
--save-bbox-dim	flag	False	Save bounding box dimensions in .txt tracks.
--save-with-object-id	flag	False	Save results with object ID in .txt files.

Results 📊

YOLOv7 Detection Only	YOLOv7 Object Tracking with ID	YOLOv7 Object Tracking with ID and Label

References

• YOLOv7 GitHub
• SORT GitHub

Some of my articles/research papers | computer vision awesome resources for learning | How do I appear to the world? 🚀

Ultralytics YOLO11: Object Detection and Instance Segmentation🤯

Parking Management using Ultralytics YOLO11

My 🖐️Computer Vision Hobby Projects that Yielded Earnings

Best Resources to Learn Computer Vision

Roadmap for Computer Vision Engineer

How did I spend 2022 in the Computer Vision Field

Domain Feature Mapping with YOLOv7 for Automated Edge-Based Pallet Racking Inspections

Exudate Regeneration for Automated Exudate Detection in Retinal Fundus Images

Feature Mapping for Rice Leaf Defect Detection Based on a Custom Convolutional Architecture

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

Explainable AI in Drug Sensitivity Prediction on Cancer Cell Lines

Train YOLOv8 on Custom Data

Session with Ultralytics Team about Computer Vision Journey

Contributors 🤝