YOLO11 Object Detection Training

This project is set up to train a YOLO11 model for object detection.

Setup

1. Create and activate a virtual environment:

   python3 -m venv .venv
   source .venv/bin/activate

2. Install dependencies:

   pip install -r requirements.txt

3. Add your data:

   • Place your training images in data/images/train
   • Place your validation images in data/images/val
   • Place your training labels in data/labels/train
   • Place your validation labels in data/labels/val

   The label files should be in YOLO format (txt files with one line per bounding box: class_id center_x center_y width height).

4. Update data/dataset.yaml:

   • Change the names list to match the classes in your dataset.

Training

To start training the model, make sure your virtual environment is activated and run the following command:

python train.py

The training results, including saved model weights, will be stored in the runs/ directory.

Docker

You can also run this project using Docker.

1. Build the Docker image:

   docker build -t cartes-yolo .

2. Run the Docker container:

   To train the model inside the container, you need to mount your local data/ and runs/ directories so that the container can access your dataset and save the training results.

   docker run -it --rm \
     -v "$(pwd)/data:/app/data" \
     -v "$(pwd)/runs:/app/runs" \
     cartes-yolo \
     --data data/dataset.yaml --epochs 100 --imgsz 640 --project runs/train --name train

   You can customize the arguments passed to train.py (e.g., --epochs, --imgsz, --name) as needed.

   Example: Running with results in a local tmp folder

   To direct the training results to a local tmp folder instead of runs, you can use the following command. This command mounts your local data directory to /data inside the container and your local tmp directory to /app/runs inside the container, which is where the training script will write its output.

   docker run -it --rm \
     -v "$(pwd)/data:/data" \
     -v "$(pwd)/tmp:/app/runs" \
     cartes-yolo \
     --epochs=1 --data=/data/dataset.yaml