Adaptive Batch Normalization (AdaBN) in PyTorch

This repository contains an implementation of Adaptive Batch Normalization (AdaBN) in PyTorch, a technique that adapts batch normalization statistics to better generalize across domain shifts and other dataset-specific variations.

Revisiting Batch Normalization For Practical Domain Adaptation

---

Files in the Repository

• batchnorm_adapt.py
  Implements the core functionality of Adaptive Batch Normalization (AdaBN). This includes adapting batch normalization statistics to new datasets or domains.

• utils.py
  Contains utility functions supporting the main implementation, such as data preprocessing, logging, and model manipulation.

---

Key Features

• Adaptation to Domain Shifts
  AdaBN updates batch normalization statistics to match the target domain, improving model generalization in scenarios like domain adaptation and test-time adaptation.

• Easy Integration
  Modular implementation to easily integrate into existing PyTorch projects.

• Reusable Utilities
  Helper functions for common operations related to adaptive normalization and data handling.

---

Getting Started

Prerequisites

• Python 3.7+
• PyTorch 1.10+
• NumPy

Install the required packages:

pip install torch numpy