huawei-noah/Efficient-AI-Backbones

Health v2 · maintenance only

Flags: no_license

Adoption not part of the score

4418 stars · 735 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

A collection of efficient neural network backbone architectures (GhostNet, TNT, ViG, WaveMLP, TinyNet, etc.) from Huawei Noah's Ark Lab, with PyTorch and MindSpore implementations and pretrained ImageNet weights. It serves as a research model zoo for lightweight and vision transformer models.

Use cases

• find efficient cnn backbones for mobile deployment
• download pretrained ghostnet weights for imagenet
• implement vision transformer models in pytorch
• compare lightweight model architectures for image classification
• use vision gnn models in my project
• compress models for edge inference

When to choose

• you need efficient, lightweight image classification backbones with pretrained weights
• you want reference implementations of published architectures like GhostNet or TNT
• you work in PyTorch or MindSpore and need vision model baselines

When to avoid

• you need a production-ready pip-installable model library with stable APIs
• you need non-vision tasks like NLP or audio
• you require a permissive license - the repo has no explicit license, limiting reuse

Facets

library · maturity active

machine-learning deep-learning image-processing deep-learning computer-vision image-processing artificial-intelligence python vision-transformer ghostnet model-zoo pretrained-models efficient-inference model-compression pytorch mindspore imagenet research-code gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="huawei-noah/Efficient-AI-Backbones")