/CapsNet-Tensorflow

A Tensorflow implementation of CapsNet(Capsules Net) in Hinton's paper Dynamic Routing Between Capsules

Primary LanguagePythonApache License 2.0Apache-2.0

CapsNet-Tensorflow

Contributions welcome License Gitter

A Tensorflow implementation of CapsNet based on Geoffrey Hinton's paper Dynamic Routing Between Capsules

capsVSneuron

Notes:

  1. The current version supports MNIST and Fashion-MNIST datasets. The current test accuracy for MNIST is 99.64%, and Fashion-MNIST 90.60%, see details in the Results section
  2. See dist_version for multi-GPU support
  3. Here(知乎) is an article explaining my understanding of the paper. It may be helpful in understanding the code.

Important:

If you need to apply CapsNet model to your own datasets or build up a new model with the basic block of CapsNet, please follow my new project CapsLayer, which is an advanced library for capsule theory, aiming to integrate capsule-relevant technologies, provide relevant analysis tools, develop related application examples, and promote the development of capsule theory. For example, you can use capsule layer block in your code easily with the API capsLayer.layers.fully_connected and capsLayer.layers.conv2d

Requirements

  • Python
  • NumPy
  • Tensorflow>=1.3
  • tqdm (for displaying training progress info)
  • scipy (for saving images)

Usage

Step 1. Download this repository with git or click the download ZIP button.

$ git clone https://github.com/naturomics/CapsNet-Tensorflow.git
$ cd CapsNet-Tensorflow

Step 2. Download MNIST or Fashion-MNIST dataset. In this step, you have two choices:

  • a) Automatic downloading with download_data.py script
$ python download_data.py   (for mnist dataset)
$ python download_data.py --dataset fashion-mnist --save_to data/fashion-mnist (for fashion-mnist dataset)
  • b) Manual downloading with wget or other tools, move and extract dataset into data/mnist or data/fashion-mnist directory, for example:
$ mkdir -p data/mnist
$ wget -c -P data/mnist http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz
$ wget -c -P data/mnist http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz
$ wget -c -P data/mnist http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz
$ wget -c -P data/mnist http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz
$ gunzip data/mnist/*.gz

Step 3. Start the training(Using the MNIST dataset by default):

$ python main.py
$ # or training for fashion-mnist dataset
$ python main.py --dataset fashion-mnist
$ # If you need to monitor the training process, open tensorboard with this command
$ tensorboard --logdir=logdir
$ # or use `tail` command on linux system
$ tail -f results/val_acc.csv

Step 4. Calculate test accuracy

$ python main.py --is_training=False
$ # for fashion-mnist dataset
$ python main.py --dataset fashion-mnist --is_training=False

Note: The default parameters of batch size is 128, and epoch 50. You may need to modify the config.py file or use command line parameters to suit your case, e.g. set batch size to 64 and do once test summary every 200 steps: python main.py --test_sum_freq=200 --batch_size=48

Results

The pictures here are plotted by tensorboard and my tool plot_acc.R

  • training loss

total_loss margin_loss reconstruction_loss

Here are the models I trained and my talk and something else:

Baidu Netdisk(password:ahjs)

  • The best val error(using reconstruction)
Routing iteration 1 3 4
val error 0.36 0.36 0.41
Paper 0.29 0.25 -

test_acc

My simple comments for capsule

  1. A new version neural unit(vector in vector out, not scalar in scalar out)
  2. The routing algorithm is similar to attention mechanism
  3. Anyway, a great potential work, a lot to be built upon

My weChat:

my_wechat

Reference