/awesome-efficient-ai-libs

A curated list of awesome AI libraries that are efficiently implemented

MIT LicenseMIT

Awesome Efficient AI Libs | Awesome

This is a curated list of awesome AI libraries which are efficiently implemented, allowing developers to train neural networks quickly.

Contents

Computer Vision

  • kaolin - A PyTorch Library for Accelerating 3D Deep Learning Research.
  • RaDe-GS/diff-gaussian-rasterization. - This is a modified version to support depth & alpha rendering (both forward and backward) which could render the normal and depth map of the 3D scene.
  • GS-IR/diff-gaussian-rasterization. - Customed rasterization algorithm for 3D Gaussian splatting, which could be used to inverse render the 3D scene.
  • r3dg-rasterization - Customed rasterization algorithm for 3D Gaussian splatting, which could be used to inverse render the 3D scene.
  • nerfacc_ray_resampling -This is the modified version of nerfacc-0.3.0's ray_resampling.
  • torch_pbr -Torch PBR is a light-weight library for differentiable PBR written purely in Python/PyTorch.
  • gaustudio - GauStudio is a modular framework that supports and accelerates research and development in the rapidly advancing field of 3D Gaussian Splatting (3DGS) and its diverse applications.
  • gsplat - This library contains the neccessary components for efficient 3D to 2D projection, sorting, and alpha compositing of gaussians and their associated backward passes for inverse rendering.
  • Differentiable Iso-Surface Extraction - This repository consists of a variety of differentiable iso-surface extraction algorithms implemented in cuda.
  • Shape As Points - Shape As Points: A Differentiable Poisson Solver for 3D Shape Reconstruction.
  • PET-NeuS/ops - Differentiable bias-activations function and 3D grid sampling in CUDA.
  • SelfReconCode/MCGpu - CUDA for differentiable marching cube algorithm.
  • SelfReconCode/MCAcc - Efficient implementation in CUDA for differentiable 3D grid sampling algorithm.
  • SelfReconCode/FastMinv - Efficient algorithm in CUDA for differentiable inverting 3x3 matrix.
  • gsplat - CUDA accelerated rasterization of gaussian splatting.
  • diff-gaussian-rasterization - Used as the rasterization engine for the paper 3DGS project.
  • pytorch3d_knn_cuda - pytorch3d_knn_cuda.
  • simple-knn - Simple KNN implementation in CUDA used in the 3DGS project.

Natural Language Processing

  • fastai - The fastai deep learning library, easy to use and fast.
  • fastcore - A fast and easy-to-use foundation library for deep learning and AI research.
  • fastnlp - A Modularized and Extensible NLP Framework. Currently still in incubation.