quick_run.md

Quick Run

Training

Run the following command to train a model:

python dettrack/detection/train_yolov8.py  ${CONFIG_FILE}

# Example
python dettrack/detection/train_yolov8.py -cfg config/yolov8n_mot_ch.yaml

Validation

Benchmark result on MOT17 dataset:

python run/val.py --data_path /home/vankhoa/datasets \
                  --yolo-model runs/detect/yolov8n_mot_ch8/weights/best.pt \
                  --tracking-method deepocsort \
                  --benchmark MOT17 \
                  --processes-per-device 4 \
                  --split val 

Inference

Simple Tracking Demo

1. Configuration: Modify the configuration file config.yaml to set your preferred detection and tracking algorithms, camera sources, and other parameters.
2. Run the Application:

python run_tracker_simple.py

Detection and Tracking Demo

python run/track.py --yolo-model runs/detect/yolov8s_crowdhuman13/weights/best.pt \
                    --tracking-method ocsort \
                    --source test2.mp4 \
                    --show \
                    --save