moskomule/senet.pytorch

Health v2 · maintenance only

Adoption not part of the score

2344 stars · 447 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

A PyTorch implementation of Squeeze-and-Excitation Networks (SENet), the ILSVRC 2017 classification winner, including SE-ResNet and SE-Inception-v3 variants. It provides trainable architectures, training scripts for CIFAR-10 and ImageNet, and pretrained models loadable via torch.hub.

Use cases

• add squeeze-and-excitation blocks to my pytorch model
• load a pretrained SE-ResNet50 with torch.hub
• train SE-ResNet on CIFAR-10
• train SENet on ImageNet with distributed GPUs
• reproduce ILSVRC 2017 winning architecture in pytorch
• compare SE-ResNet accuracy against plain ResNet

When to choose

• you need SENet/SE-ResNet architectures in PyTorch with pretrained weights
• you want a lightweight, MIT-licensed reference implementation for research
• you want models loadable via torch.hub

When to avoid

• you need actively maintained code with recent PyTorch support (last release 2021)
• you want a broad model zoo rather than SENet specifically
• you need production training pipelines with configurable hyperparameters

Facets

library · maturity maintenance

machine-learning deep-learning deep-learning computer-vision image-processing python pytorch senet squeeze-and-excitation image-classification torch-hub pretrained-models resnet inception gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="moskomule/senet.pytorch")