Finger Recognition Using CNN in TFLearn
TFlearn is a modular and transparent deep learning library built on top of Tensorflow. It was designed to provide a higher-level API to TensorFlow in order to facilitate and speed-up experimentations, while remaining fully transparent and compatible with it.
link: http://tflearn.org/
I have used TFLearn to create CNN and openCV to collect and store data, and to process data.
- record.py uses openCV to record and store data in folders.
- Press 1,2,3,4 or 5 to change the finger image that is being stored.
- press 'space' to store the image.
- press 'esc' to quit.
- Use train.py to train the CNN. It automatically saves the weights when it finishes training.
- Use run_cnn.py to run the finger detection. It automatically loads the saved weights.
元リポジトリのをpython3.7に対応させました。