Use guide

Create environment

This project requires Python 3.10 or above.

1. Create a virtual environment

python3 -m venv .venv

2. Activate the environment

source ./.venv/bin/activate

3. Install packages

pip install -r requirements.txt

Download dataset

1. Run the following command to download the dataset.

python3 download_dataset.py

2. You can see the visualization of sample images from the dataset by running the following command

python3 plot_dataset.py

Train model

Run the following commands to train each model

# Nano model
python3 train_nano.py

# Small model
python3 train_small.py

# Medium model
python3 train_medium.py

Evaluate model

Run the following command to evaluate the model

python3 model_evaluation.py

Run inference on examples

Video file

General command:

python3 process_video.py --input <path/to/input/file> --model <path/to/model> --output <path/to/output/folder>

Example usages:

# Yolov11 nano model
python3 process_video.py --input video_input/warsaw.mp4 --model runs/detect/yolo11n_b64_e50/weights/bests.pt

# Yolov11 small model
python3 process_video.py --input video_input/warsaw.mp4 --model runs/detect/yolo11s_b32_e50/weights/best.pt

# Yolov11 medium model
python3 process_video.py --input video_input/warsaw.mp4 --model runs/detect/yolo11m_b16_e50/weights/best.pt

Camera feed

General command:

python3 process_camera.py --model <path/to/model>

Example usages:

# Yolov11 nano model
python3 process_camera.py --model runs/detect/yolo11n_b64_e50/weights/best.pt

# Yolov11 small model
python3 process_camera.py --model runs/detect/yolo11s_b32_e50/weights/best.pt

# Yolov11 medium model
python3 process_camera.py --model runs/detect/yolo11m_b16_e50/weights/best.pt