---

ShiftedBronzes Benchmark and OOD Detection Codebase

This repository provides the official implementation of our paper:
"ShiftedBronzes: Benchmarking and Analysis of Domain Fine-grained Out-of-Distribution Detection in Gradual Shifts".

We provide:

• Training and evaluation code for ID-like and LoCoOp under different background settings.
• Code for testing post-hoc OOD detection methods adapted from OpenOOD.

---

📁 Project Structure

ShiftedBronzes/

├── ID-like-train/               # Training/testing code for ID-like using original background

├── ID-like-train-change-bg/    # Training/testing code for ID-like using replaced background

├── LoCoOp-train/               # Training/testing code for LoCoOp using original background

├── LoCoOp-train-change-bg/     # Training/testing code for LoCoOp using replaced background

├── OpenOOD/                    # Post-hoc OOD detection testing code (based on OpenOOD)

├── zest\_code/                  # Zest code for generating transferred data

├── requirements.txt            # Environment requirements

└── README.md                   # Project README

---

🔧 Installation

git clone https://github.com/your_username/ShiftedBronzes.git
cd ShiftedBronzes
pip install -r requirements.txt

---

🗂 Dataset

You can download the ShiftedBronzes dataset in this Link.

---

Training & Evaluation

We provide scripts to train and test each models under corresponding code folder.