ozora-ogino / efficient_backbones

Implementation of efficient backbones for computer vision task.

Efficient Backbones

Efficient backbones is the project which aims to provide efficient SOTA vision backbones based on PyTorch.

Currently following models are supported;

• Resnet
• EfficientNetv2
• Xception
• Ghostnet
• G-Ghost RegNet
• LambdaNet
• HarDNet

NOTE: All models implemented in this project doesn't have head.

Installation

pip install git+https://github.com/ozora-ogino/efficient_backbones

Usage

All model have out_channels attribute in order to access the number of out channels of backbone in the same way.

from efficient_backbones import build_backbone

resnet_backbone = build_backbone("resnet18")
out_channels = resnet_backbone.out_channels