YOLO Dataset for Students' Behavior Detection

Steps

• Data collection and recording
• Convert videos into frames
• Data Labelling & Annotation
• Convert Json format into txt format
• Splitting images and labels into train, val and test folders
• Model Training
• Detection Results

Implementation

• Create & Activate Conda Environment

conda create –n yolov5 python=3.8
conda activate yolov5

• Clone YOLOv5 & Install Requirments

git clone https://github.com/ultralytics/yolov5.git
pip install -r requirements.txt

Training Custom Model (Students' Behaviour Detection )

• Model training using pre-trained model (yolov5s.pt) model on GPU (device 0)

python train.py --img 640 --batch 16 --epochs 100 --data data.yaml --weights yolov5s.pt --device 0

Exporting Model into ONNX

python export.py --weights runs/train/exp/weights/best.pt --include onnx

Inference

python detect.py --weights runs/train/exp/weights/best.pt --source demo.mp4 --conf 0.25

References

• https://docs.ultralytics.com/
• https://docs.ultralytics.com/yolov5/quickstart_tutorial/
• https://github.com/ultralytics/yolov5