Lightweight pytorch functions for neural network featuremap sampling.

WARNING: API is not yet stable. API subject to change!

Introduction

Sampling neural network featuremaps at explicit coordinates has become more and more common with popular developments like:

• Learning Continuous Image Representation with Local Implicit Image Function

• NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis

• PointRend: Image Segmentation as Rendering

PyTorch provides the tools necessary that to sample coordinates, but they result in a large amount of error-prone code. TorchSample intends to make it simple so you can focus on other parts of the model.

Contents:

• Installation
• Overview
• Generating Coordinates
  • Random Coordinates
  • Comprehensive Coordinates
  • Helpers
• Sampling
  • Sampling with positional encoding
• No Batch
• Models
  • MLP
• Positional Encoding
• About Align Corners
  • History
  • Explanation
• API Reference
  • torchsample package
• Changelog
  • Version 0.1.0
• Module Index
• Index