rickcrawford / cards_dataset

This is a training dataset with Bicycle Jumbo playing cards.

Cards Image Dataset

This is a card detection dataset built for Bicycle Jumbo playing cards for Darknet YOLOv3. This is heavily based off of the work done in geaxgx/playing-card-detection to generate the training set. I have based the notebook in this project based on the work done in the notebook avaiable in the project above.

This dataset includes cards that are labeled from [2-A] for the 4 suits [c,d,h,s], plus 2 jokers JK

Based this dataset off of the Racoons Dataset

The create_validation_set.sh file is from experiencor/keras-yolo3

Create a Virtual Python environment:

python3 -m venv .venv
source .venv/bin/activate
pip install --upgrade pip
pip install -r requirements.txt

Run the notebook:

jupyter notebook

To build the dataset using YOLO v3 Darknet, first download the darknet53.conv.74 file, and then kick off the training based on the instructions from the yolo website.

darknet detector train data/obj.data cfg/yolov3.cfg darknet53.conv.74 -map