OPEN SOURCE

Torch Examples, Guides and Resources

• Outline
• Torch Examples
• Custom Torch Modules in C/CUDA
• Variational Auto-Encoders
• PointNet
• Volumetric Nearest Neighbor Upsampling

This page presents a repository including beginner-friendly as well as advanced Torch examples, a template for implementing custom Torch modules in C/CUDA and multiple variants of variational auto-encoders as well as an implementation of PointNet. The code is accompanies by a series of blog articles:

• Examples for Getting Started with Torch for Deep Learning
• More Examples for Working with Torch
• Implementing Torch Modules in C/CUDA
• Variational Auto-Encoder in Torch
• Denoising Variational Auto-Encoder in Torch
• Bernoulli Variational Auto-Encoder in Torch
• PointNet Auto-Encoder in Torch

Torch Examples

This part of the repository includes examples for getting started with Torch, ranging from training simple auto-encoders, to fine-tuning, handling weight initialization and saving/loading models:

Custom Torch Modules

Torch can easily be extended using custom modules implemented in C/CUDA. This part of the repository includes a simple example and template for implementing custom modules:

Variational Auto-Encoders

This part of the repository includes implementations of variational auto-encoders (VAEs) [1], denoising VAEs [2] and categorical/Bernoulli VAEs [3,4]:

PointNet

This part includes an implementation of PointNet, inspired by [5,6]:

Volumetric Nearest Neighbor Upsampling

A torch module for volumetric nearest neighbor upsampling for 3D neural networks:

• [1] D. P. Kingma, S. Mohamed, D. J. Rezende, and M. Welling. Semisupervised learning with deep generative models. In Advances in Neural Information Processing Systems, pages 3581–3589, 2014.
• [2] D. J. Im, S. Ahn, R. Memisevic, and Y. Bengio. Denoising criterion for variational auto-encoding framework. In AAAI Conference on Artificial Intelligence, pages 2059–2065, 2017.
• [3] E. Jang, S. Gu, and B. Poole. Categorical reparameterization with gumbel-softmax. CoRR, abs/1611.01144, 2016.
• [4] C. J. Maddison, A. Mnih, and Y. W. Teh. The concrete distribution: A continuous relaxation of discrete random variables. CoRR, abs/1611.00712, 2016.
• [5] Charles Ruizhongtai Qi, Hao Su, Kaichun Mo, Leonidas J. Guibas: PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation. CoRR abs/1612.00593 (2016)
• [6] Haoqiang Fan, Hao Su, Leonidas J. Guibas: A Point Set Generation Network for 3D Object Reconstruction from a Single Image. CoRR abs/1612.00603 (2016)