A modular deep learning system for data handling and classification
the main components:
- Data hanndler - load the data, split it to train\validation, and create random batches for the classifier to train on
- Classifier - initialize the tensorflow graph, uses modular components for net architechture, cost function and solver.
The system can easily be converted for other applications (more interesting than Cifar10)
Full cifar10 data set can be found here: https://www.kaggle.com/c/cifar-10/data