YOLO Object Detection for Playing Cards

Training a YOLOv2 model with the Darknet object detection framework for real-time detection of playing cards

Demo

Click the video below:

I ran YOLOv3 on my own generated dataset of 60,000 images (20% saved as test set). Results:

mAP	IoU	F1-Score
99.86%	82.92%	0.99

The script runs inference at real-time (30fps) on a live video feed.

Only tested on Bicycle cards. I chose this deck due to its popularity in commercial use.

Name		Name	Last commit message	Last commit date
Latest commit History 4 Commits
.ipynb_checkpoints		.ipynb_checkpoints
.gitignore		.gitignore
Generate Poker Hands Dataset.ipynb		Generate Poker Hands Dataset.ipynb
Poker Hands Model Training.ipynb		Poker Hands Model Training.ipynb
README.md		README.md
Utils.ipynb		Utils.ipynb
Utils.py		Utils.py
bash script.txt		bash script.txt
convert_voc_yolo.py		convert_voc_yolo.py
data_generation.py		data_generation.py
poker-data-css-style.css		poker-data-css-style.css
poker_rf_optimized.PNG		poker_rf_optimized.PNG