SHC-IR

The code of paper "Image Retrieval with Well-Separated Semantic Hash Centers" accepted by ACCV 2022

Environment

• PyTorch 1.7.1 + cuda 11.2

Get Started

1. three dataset ImageNet, Stanford Cars, NABirds should be download from their official website
2. run main_cls.py to generate and save semantic category for three datasets.
3. run optimizeAccel.py to generate and save hash centers for each datasets.
4. run main.py to get mAP results for three datasets.

Note

the parameter setting can be modified in scripts\utils.py
some setting should be modified in scripts\head.py to fit for generate semantic category or retrieval