Efficient-AI-Backbones  by huawei-noah

Efficient AI backbones for computer vision research

created 5 years ago
4,270 stars

Top 11.7% on sourcepulse

GitHubView on GitHub
Project Summary

This repository provides a collection of efficient AI backbones for computer vision tasks, developed by Huawei Noah's Ark Lab. It targets researchers and practitioners seeking high-performance, computationally inexpensive models, offering a range of architectures like GhostNet, TNT, and MLP variants.

How It Works

The project implements novel architectural designs that prioritize computational efficiency. Key innovations include "Ghost Modules" which generate more features from cheap operations, and hybrid approaches combining convolutional and transformer elements (e.g., CMT, TNT). This strategy aims to reduce FLOPs and parameter counts while maintaining or improving accuracy, making models suitable for resource-constrained environments.

Quick Start & Requirements

  • Installation typically involves cloning the repository and installing dependencies via pip.
  • Requires Python and PyTorch. Specific model implementations may have additional dependencies detailed in their respective subdirectories.
  • Links to official quick-start guides and documentation are not explicitly provided in the README, but individual model directories contain PyTorch code.

Highlighted Details

  • Includes implementations of GhostNet, GhostNetV2, TNT (Transformer in Transformer), AugViT, WaveMLP, and ViG.
  • Several models have received accolades, such as GhostNet and TNT being recognized as "Most Influential" papers at CVPR 2020 and NeurIPS 2021, respectively.
  • Offers both PyTorch and MindSpore implementations for many models.
  • Features models addressing parameter efficiency (ParameterNet) and brain-inspired computing (SNN-MLP).

Maintenance & Community

  • Developed by Huawei Noah's Ark Lab.
  • Recent activity includes the release of GhostNetV2 and G-GhostNet code in late 2022, and ParameterNet accepted to CVPR 2024.
  • No explicit community links (Discord, Slack) are provided in the README.

Licensing & Compatibility

  • The README does not specify a license.
  • Compatibility for commercial use or closed-source linking is undetermined without a license.

Limitations & Caveats

The README does not specify a license, which is a significant blocker for assessing commercial usability or derivative works. Documentation for individual models appears to be within their respective subdirectories, requiring users to navigate them separately.

Health Check
Last commit

4 months ago

Responsiveness

1 day

Pull Requests (30d)
0
Issues (30d)
0
Star History
77 stars in the last 90 days

Explore Similar Projects

Feedback? Help us improve.