Prerequisites

• Python == 3.8, numpy == 1.20.3, faiss

Datasets

The tested datasets are available at https://www.cse.cuhk.edu.hk/systems/hash/gqr/datasets.html.

1. Download the dataset GIST from ftp://ftp.irisa.fr/local/texmex/corpus/gist.tar.gz. The dataset is large. This step may take several minutes. The data format can be found in http://corpus-texmex.irisa.fr/.

   wget -O ./gist.tar.gz ftp://ftp.irisa.fr/local/texmex/corpus/gist.tar.gz --no-check-certificate

2. Unzip the dataset.

   tar -zxvf ./gist.tar.gz -C ./

3. Preprocess the dataset with random orthogonal transformation.

   python randomize.py

4. Generate the clustering of the dataset for IVF.

   python ivf.py