Home / Python / Deep Learning

Pytorch easy-to-follow Capsule Network tutorial

Capsule-Network-Tutorial This is easy-to-follow Capsule Network tutorial with clean readable code: Capsule Network.ipynb Dynamic Routing Between Capsules Understanding Hinton’s Capsule Networks blog posts: Part I: Intuition.
Information
Category: Python / Deep Learning
Watchers: 11
Star: 748
Fork: 135
Last update: Aug 14, 2022
Resource links
https://github.com/higgsfield/Capsule-Network-Tutorial
12 Open More issues
Closed

Class PrimaryCaps initialization parameter wrong?

class PrimaryCaps(nn.Module): def __init__(self, num_capsules=8, in_channels=256, out_channels=32, kernel_size=9):

Per paper: "The second layer (PrimaryCapsules) is a convolutional capsule layer with 32 channels of convolutional 8D capsules (i.e. each primary capsule contains 8 convolutional units with a 9 × 9 kernel and a stride of 2)"

Which indicate num_capsules should be 32 and out_channels should be 8.

opened Oct 4, 2019 by donglinz 1
Closed

MLP Capsule Examples?

Awesome repo. I've been trying to implement an MLP based caps net, where the primary caps analyzes subsets of features in groups, feeding them into MLPs per capsule then pushed to downstream capsules.

For some reason, I am getting terrible results. Do you know what I might be doing wrong?

What are areas that make MLP capsules different from convolutional primary capsules, what should I be aware of? How do I optimize to get good results. I think dynamic routing is essential for establishment of hierarchy, but in the non-image context, I'm unsure of how the affine transformation is helpful.

Can you post code on how one might do this using MLP instead of convolutional layers?

opened Aug 7, 2019 by jlevy44 1
Closed

Fix initialization parameters in PrimaryCaps class

Fix issue 13 https://github.com/higgsfield/Capsule-Network-Tutorial/issues/13

opened Oct 4, 2019 by donglinz 0
Closed

I can not make it work with CIFAR10

I wanted to try working on CIFAR10, I modified the channel number =3 in convLayer e the kernel dim = 24 in convLayer. But it is not working in the primaryCaps u = u.view(x.size(0), 32 * 6 * 6, -1) that is giving to me errot :

ipython-input-4-7b4e2b87bd5c> in forward(self, x) 14 print( "PrimaryCaps {}".format(x.size())) 15 #u = u.view(x.size(0), 32 * 6 * 6, -1) ---> 16 u = u.view(x.size(0), 32 * 6 * 6, -1) 17 return self.squash(u) 18 RuntimeError: invalid argument 2: size '[100 x 1152 x -1]' is invalid for input with 204800 elements at /pytorch/aten/src/TH/THStorage.c:37

Do you have any advise for me?

opened Feb 13, 2019 by Riretta 0
Open

weight are nan

hello,I copy the code in the pycharm,but I found that the weight are 'nan',why?

opened Jan 29, 2021 by Sample-design-alt 0
Open

TRY MY IMPLEMENTATION

Similar to others, I found many issues with this implementation (a lot of mistakes!!!). So I decided to create my own one. It is bug-free and works very good. You can find it here:

https://github.com/hula-ai/capsule_network_dynamic_routing

opened Nov 21, 2019 by amobiny 0
Open

Use tensor.item() for scalar conversion

To handle

IndexError:_ invalid index of a 0-dim tensor. Use tensor.item() to convert a 0-dim tensor to a Python number

opened May 13, 2019 by ketankr9 0
Open

test accuracy

I don't know why the train accuracy is becoming lower and lower with the epoch decreasing. Besides, the margin loss and reconstruction loss can't be caculated.

opened Mar 17, 2019 by yimzhai3 0
Open

Update Capsule Network.ipynb

In the latest version of Pytorch loss.data returns the value, loss.data[0] throws an error.

opened Dec 27, 2018 by kunalmessi10 0
Open

Softmax in routing algorithm incorrect?

Hi, I think the softmax in the routing algorithm is being calculated over the wrong dimension.

Currently the code has:

b_ij = Variable(torch.zeros(1, self.num_routes, self.num_capsules, 1))
       ...
        for iteration in range(num_iterations):
            c_ij = F.softmax(b_ij)

and since the dim parameter is not passed to the F.softmax call it will choose dim=1 and compute the softmax over the self.num_routes dimension (input caps or 1152 here) whereas the softmax should be computed so that the c_ij between each input capsule and all the capsules in the next layer should sum to 1.

Thus the correct call should be:

           c_ij = F.softmax(b_ij, dim=2)
opened May 17, 2018 by geefer 1

Contents

    bourdakos1 A Tensorflow implementation of Capsule Networks
    JunYeopLee A Tensorflow(v1.4) implementation of Capsule Networks (Dynamic Routing Between Capsules , https://arxiv.org/abs/1710.09829)
    cedrickchee [NO MAINTENANCE INTENDED] A PyTorch implementation of CapsNet architecture in the NIPS 2017 paper "Dynamic Routing Between Capsules".
    naturomics CapsLayer: An advanced library for capsule theory
    sekwiatkowski A curated list of awesome resources related to capsule networks
    yongheng1991 3D Point Capsule Networks
    ZhenchaoTang Re-implementation of the vector capsule with dynamic routing
    puniverse Dead-Simple Packaging and Deployment for JVM Apps
    capsidjs TodoMVC implementation in Capsule. 3.45 kB gzipped
    vlraik An experimental Generative Adversarial Network implementation of the Capsule Neural Network
    daiquocnguyen A Capsule Network-based Model for Learning Node Embeddings
    EscVM Official TensorFlow code for the forthcoming paper "Efficient-CapsNet: Capsule Network with Self-Attention Routing".
    omerbsezer Pytorch Tutorial, Pytorch with Google Colab, Pytorch Implementations: CNN, RNN, DCGAN, Transfer Learning, Chatbot, Pytorch Sample Codes
    jupyterlab Tutorial material and instruction for scipy 2018 jupyterlab tutorial
    Eurus-Holmes :crystal_ball: Life is short, you need PyTorch.
    jsikyoon Tensorflow Implementation of Adversarial Attack to Capsule Networks
    mbertschler Autopilot in Go for docking the SpaceX Dragon capsule in the simulator
    r-spacex SpaceX REST API - 🚀 Open Source REST API for rocket, core, capsule, pad, and launch data
    klinker24 A light-weight, easy-to-implement, and easy-to-look-at way to do a tutorial pager or dialog on Android
    lesteve scikit-learn tutorial
    yunjey PyTorch Tutorial for Deep Learning Researchers
    Share this repo

    Related Repos


    Deep Learning
    837
    An implementation of chunked, compressed, N-dimensional arrays for Python.
    zarr-developers Zarr Latest Release Package Status License Build Status Coverage Downloads Gitter Citation What is it? Zarr is a Python package providing an implement
     
    Deep Learning
    317
    Serverless proxy for Spark cluster
    Hydrospheredata Hydrosphere Mist Hydrosphere Mist is a serverless proxy for Spark cluster. Mist provides a new functional programming framework and deployment model f
     
    Deep Learning
    2.5k
    Distributed Tensorflow, Keras and PyTorch on Apache Spark/Flink & Ray
    intel-analytics Distributed TensorFlow, PyTorch, Keras and BigDL on Apache Spark & Ray Analytics Zoo is an open source Big Data AI platform, and includes the followin
     
    Deep Learning
    10
    Black-Scholes StarkNet Library
    araghava Cairo Black-Scholes Library Black-Scholes library implemented as a Cairo smart contract. All inputs, outputs, and internal calculations use 27-digit f
     
    Deep Learning
    5
    Motoko library to help create an append-only logger actor
    ninegua IC Logger This motoko library provides a module to help create an append-only logger actor. Usage You can use this library with the vessel package man
     
    Deep Learning
    4
    LeetCode_Solution
    ashishpatel26 LeetCode_Solution LeetCode: https://leetcode.com/problemset/all/ # Title Difficulty Colab Code 1 Two Sum Easy 2 Add Two Numbers Medium 3 Longest Subst
     
    Deep Learning
    4
    OCR-D wrapper for detectron2 based segmentation models
    bertsky ocrd_detectron2 OCR-D wrapper for detectron2 based segmentation models Introduction Installation Usage OCR-D processor interface ocrd-detectron2-segm
     
    Deep Learning
    134
    Check for LDAP protections regarding the relay of NTLM authentication
    zyn3rgy LDAP Relay Scan A tool to check Domain Controllers for LDAP server protections regarding the relay of NTLM authentication. If you're interested in the
     
    Deep Learning
    66
    Fuzzware's main repository. Start here to install.
    fuzzware-fuzzer Fuzzware Fuzzware is a project for automated, self-configuring fuzzing of firmware images. The idea of this project is to configure the memory ranges
     
    Deep Learning
    35
    Time Weighted Asset Mints
    abigger87 TWAM • Time Weighted Asset Mints A minting harness enabling time-weighted assets to determine minting prices. How it works Requirements: maxMintingAmo
     
    Deep Learning
    33
    A complete, self-contained example for training ImageNet at state-of-the-art speed with FF...
    libffcv ffcv ImageNet Training A minimal, single-file PyTorch ImageNet training script designed for hackability. Run train_imagenet.py to get... ...high accur
     
    Deep Learning
    33
    เอกสารประกอบการสอนทุกเนื้อหาในช่องยูทูป KongRuksiam Official
    kongruksiamza 📖 เอกสารประกอบการสอน (PDF) 🔏 Blockchain & Smart Contract Blockchain เบื้องต้น Smart Contract & Solidity เบื้องต้น 💻 Programming เขียนโปรแกรมภาษา Ja
     
    Deep Learning
    34
    Omnivore A Single Model for Many Visual Modalities
    facebookresearch Omnivore: A Single Model for Many Visual Modalities [paper][website] OMNIVORE is a single vision model for many different visual modalities. It learns
     
    Deep Learning
    30
    Python implementation of the Lox language from Robert Nystrom's Crafting Interpreters
    dabeaz pylox Python implementation of the Lox language from Robert Nystrom's Crafting Interpreters. https://craftinginterpreters.com. This only implements th
     
    Deep Learning
    29
    ParaGen is a PyTorch deep learning framework for parallel sequence generation.
    bytedance ParaGen ParaGen is a PyTorch deep learning framework for parallel sequence generation. Apart from sequence generation, ParaGen also enhances various N