Home

Redner is a differentiable renderer. It takes a 3D scene, including geometry, materials, camera, light sources, represented by PyTorch/TensorFlow tensors, and outputs an image, also represented as a PyTorch/TensorFlow tensor. It provides necessary machinery for correctly propagating the gradients of the output image to the scene parameters. For the theory behind redner, please consult our paper "Differentiable Monte Carlo Ray Tracing through Edge Sampling". This page is a tutorial for using redner. It is still work in progress. Please let us know what to improve through email (tzumao@mit.edu) or Github issues.

Installation

New Tutorials with Google Colab

Hello redner!

How to load an object and render it in redner.
PyTorch TensorFlow

Pose estimation

How to optimize the pose of an object using redner.
PyTorch TensorFlow

Fast local shading

How to do local lighting using deferred rendering in redner.
PyTorch TensorFlow

Background images and color spaces

How to blend the rendering output with a background image, and why you should do it in the linear color space.
PyTorch TensorFlow

Physically-based rendering in redner

How to do physically-based rendering using path tracing in redner.
PyTorch TensorFlow

Materials, textures, and lighting in redner

Redner's material and light source models.
PyTorch TensorFlow

Camera models in redner

Redner's camera models.
PyTorch TensorFlow

Batch rendering

Batch rendering in redner.
PyTorch TensorFlow

Documentation

https://redner.readthedocs.io/en/latest/

Old Tutorials

Tutorial 1: optimizing a single triangle

Tutorial 2: pose estimation

Tutorial 3: materials and textures (WIP)

Tutorial 4: fast deferred rendering

Tutorial 5: coarse-to-fine estimation

Tutorial 6: adversarial examples (WIP)