# MNIST-Train Data For YOLO

This software generats MNIST-Train Data For YOLO.

Installation

Install Keras

This software using Keras.

• Keras installation instructions.

If you want to run without to think keras and backend deeplearning frameworks, enter this command.

pip install tensorflow
pip install keras

Install Darknet YOLO

git clone https://github.com/pjreddie/darknet
cd darknet
make

• YOLO: Real-Time Object Detection.

How to use

Generate MNIST Train Data

1. generat MNIST images and labels.

python mnist_to_jpg_and_label.py

2. generate train.txt and test.txt

   python generate_train_txt_and_test_txt.py

Training YOLO on MNIST

3. Copy files to darknet

   cp cfg/tiny-yolo-mnist.cfg <darknet_dir>/cfg
   cp cfg/voc-mnist.data <darknet_dir>/cfg
   cp data/voc-mnist.names <darknet_dir>/data

4. Modify train and test data path. Edit /cfg/voc-mnist.data

   train  = <path-to-mnist-train>/train.txt
   valid  = <path-to-mnist-test>/test.txt

5. Download Pretrained Convolutional Weights

   cd <darknet_dir>
   wget https://pjreddie.com/media/files/darknet19_448.conv.23

6. Make directory to save trained model.

   mkdir backup

7. Train The Model

   ./darknet detector train cfg/voc-mnist.data cfg/tiny-yolo-mnist.cfg darknet19_448.conv.23

   Predict MNIST test data

./darknet detector test <data file> <cfg file> <weights> <predict image>  

ex. command.

./darknet detector test cfg/voc-mnist.data cfg/tiny-yolo-mnist.cfg weights/tiny-yolo-mnist_500000.weights ~/MNIST-TrainDataForYOLO/JPEGImages/60015.jpg

MNIST Trained Weights

weights/tiny-yolo-mnist_500000.weights

License

MIT

Link

YOLO: Real-Time Object Detection