ResNet-DCBAM

A clean, GitHub-ready PyTorch implementation of DCBAM (Dual Channel + Spatial Attention) and a ResNet50 backbone with optional DCBAM integration. This repository contains typed, well-documented modules, an example, packaging files, and a clear README.

---

Project layout

resnet_dcbam/
├─ LICENSE
├─ README.md (this file)
├─ requirements.txt
├─ pyproject.toml
├─ .gitignore
├─ models/
│  ├─ __init__.py
│  ├─ attention/
│  │  └─ dcbam.py
│  └─ backbone/
│     └─ resnet_dcbam.py
├─ examples/
│  └─ test_backbone.py
└─ scripts/
   └─ package_check.py

---

Quick install

python -m venv .venv
source .venv/bin/activate  # or .venv\Scripts\activate on Windows
pip install -r requirements.txt

---

Design decisions

• Modular structure: models.attention.dcbam is independent and reusable.
• Backbone: models.backbone.resnet_dcbam encapsulates the ResNet50-like backbone and uses pretrained weights where possible.
• Typing & docstrings: all public functions/classes include type hints and clear docstrings.
• Tests / examples: examples/test_backbone.py demonstrates usage and verifies shapes.
• License: MIT.

---

Files (full contents)

See the models/, examples/ and scripts/ folders for implementation and usage examples.