This projects aims at classifying clothes among 9 classes using a convolutional neural network (CNN).

The implementation is all done using Google's deep learning library TensorFlow.

Prior to running the scripts, please make sure that you have installed the requirements.txt within a virtualenv.

Also, set up accordingly the paths of the Tianchi dataset in Loader.py (Loader.data_dir), train.py (FLAGS.tianchi), eval.py and app.py.

preprocess folder contains the source files used to preprocess the data (from RGB to grayscale and normalisation step).

In order to train the scripts, type python cnn/eval.py.

In order to evaluate the latest trained model, type python cnn/train.py.

More compact, you can just run sh launch.sh and add the additional flags you want to set.

Trained models are saved in the checkpoints/ folder. The hyper_params.json file contains the hyper-parameters used to train the model saved.

The summaries used by Tensorboard are saved in model/logs/summaries folder.

In order to visualise the results in Tensorboard, type within the TensorFlow environnement, tensorboard --logdir=logs/.

The file cnn/app.py is a small app that restore the model stored in model/checkpoints and test the data contained in test_app/images/.