Large-scale Image Retrieval using Textual Search Engines

The code implements the techniques presented in the following papers:

Amato, G., Bolettieri, P., Carrara, F., Falchi, F. and Gennaro, C., 2018, June. Large-Scale Image Retrieval with Elasticsearch. In The 41st International ACM SIGIR Conference on Research & Development in Information Retrieval (pp. 925-928). ACM.

Amato, G., Carrara, F., Falchi, F., Gennaro, C. and Vadicamo, L., 2019. Large-scale instance-level image retrieval. Information Processing & Management, p.102100.

Main files

• baseline.py: non-approximated bruteforce matching using original features
• scalar_quantization.py: thresholded scalar quantization in main memory
• deep_permutation_basics.py: example for computing CReLU-ed permutations
• ivfpq.py: main-memory product-quantized inverted files (using FAISS)
• sq_lucene.py: implements thresholded scalar quantization in Apache Lucene (see lucene/Dockerfile)
• sq_elastic.py: implements thresholded scalar quantization in Elasticsearch (see elasticsearch/setup_elastic.sh)