Pokemon object detections

Using yolov8 after training with Google Colab

Dataset

Because of lack of data, there are only 7 classes: pikachu, charmander, bulbasaur, squirtle, eevee, jigglypuff and other.

Requirements

pip install ultralytics

Preprocess Data

The convert.py used to convert .xml label file to .txt yolo label file.
Run resize_image.py to resize image's width to 640.

Train with Colab

Edit name.yaml. Upload images and labels.

!yolo train model=yolov8n.pt data=/content/name.yaml epochs=50 imgsz=640

Training's Result

last.pt and best.pt in result folder.

Predict

Run predict.py to see result. This is my predict to test.mp4:

Thank you for stopping by!