A Pytorch implementation of CapsNet in the paper:
Sara Sabour, Nicholas Frosst, Geoffrey E Hinton. Dynamic Routing Between Capsules. NIPS 2017
Thanks for https://github.com/XifengGuo/CapsNet-Pytorch
TODO
- Conduct experiments on other datasets.
- Explore interesting characteristics of CapsuleNet.
- Test the reconstruction results on the EMNIST-letters
Usage: Same as the https://github.com/XifengGuo/CapsNet-Pytorch
Step 1. Train a CapsNet on MNIST
Training with default settings:
python capsulenet.py
Launching the following command for detailed usage:
python capsulenet.py -h
Step 2. Test model and show reconstruction results
Suppose you have trained a model using the above command, then the trained model will be
saved to result/trained_model.pkl. Now just launch the following command to get test results.
python capsulenet.py --testing --weights result/trained_model.pkl
It will output the testing accuracy and show the reconstructed images
Reconstruction result
Digits at top 5 rows are real images from EMNIST and digits at bottom are corresponding reconstructed images.
All the results are based on 5 epochs traing. Time for training is 480s/epoch on GTX1060
Results can showing by both one-channel and color
