CapsNet-pytorch and capsule-networks
These are **competitors** — both are independent PyTorch implementations of the same foundational capsule network architecture from the same paper, and users would typically choose one based on code quality, documentation, or community preference rather than use them together.
About CapsNet-pytorch
adambielski/CapsNet-pytorch
PyTorch implementation of NIPS 2017 paper Dynamic Routing Between Capsules
About capsule-networks
gram-ai/capsule-networks
A PyTorch implementation of the NIPS 2017 paper "Dynamic Routing Between Capsules".
Implements the iterative routing-by-agreement mechanism where lower-level capsules route outputs to higher-level capsules based on prediction agreement, enabling the network to learn instantiation parameters (pose, scale, rotation) rather than just class probabilities. Integrates with TorchNet for training, TorchVision for MNIST preprocessing, and Visdom for real-time visualization, achieving 99.7% accuracy on MNIST with configurable routing iterations and batch sizes.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work