Clothing Classification with Capsule Networks

Keras (Backend: TF) implementation of FashionCapsNet

Accepted to International Journal of Informatics Technologies. Paper will be released soon!

Reference: Sara Sabour, Nicholas Frosst, Geoffrey E Hinton. Dynamic Routing Between Capsules. NIPS 2017

Base code for Capsule architecture: XifengGuo

Dataset: DeepFashion contains 290K training, 40K validation and 40K test images with 46 fine-grained category labels for clothing images

Please feel free to open an issue or to send an e-mail to furkan.kinli@ozyegin.edu.tr

• NumPy 1.16.X
• Tensorflow 1.12.X
• Keras 2.2.X
• OpenCV 3.4.X

python main.py --args

Validation accuracy converges at 255. epoch.

Apprx. 15 days to complete train on 2 GTX1080Tis.

For running with different parameters, please view the config file!

python main.py -t -w ./result/t_model.h5

Test accuracy:

• Top-1: 63.61%
• Top-3: 83.18%
• Top-5: 89.83%