Approximate search with quantized sparse representations - Archive ouverte HAL Access content directly
Conference Papers Year : 2016

Approximate search with quantized sparse representations

Abstract

This paper tackles the task of storing a large collection of vectors, such as visual descriptors, and of searching in it. To this end, we propose to approximate database vectors by constrained sparse coding, where possible atom weights are restricted to belong to a finite subset. This formulation encompasses, as particular cases, previous state-of-the-art methods such as product or residual quantization. As opposed to traditional sparse coding methods, quantized sparse coding includes memory usage as a design constraint, thereby allowing us to index a large collection such as the BIGANN billion-sized benchmark. Our experiments , carried out on standard benchmarks, show that our formulation leads to competitive solutions when considering different trade-offs between learning/coding time, index size and search quality.
Fichier principal
Vignette du fichier
paper_HAL.pdf (443.61 Ko) Télécharger le fichier
QaPQ_recall_gist1m.pdf (9.28 Ko) Télécharger le fichier
QaPQ_recall_sift1m.pdf (9.18 Ko) Télécharger le fichier
QaPQ_recall_vlad500k.pdf (9.24 Ko) Télécharger le fichier
QaRVQ_recall_gist1m.pdf (9.2 Ko) Télécharger le fichier
QaRVQ_recall_vlad500k.pdf (9.17 Ko) Télécharger le fichier
bigann_results.pdf (18.58 Ko) Télécharger le fichier
distortion_gist1m.pdf (8.77 Ko) Télécharger le fichier
distortion_sift1m.pdf (8.63 Ko) Télécharger le fichier
distortion_vlad500k.pdf (8.65 Ko) Télécharger le fichier
l2_recall_64_gist1m_CQ.pdf (8.59 Ko) Télécharger le fichier
l2_recall_64_sift1m_AQCQ.pdf (8.96 Ko) Télécharger le fichier
l2_recall_72_gist1m.pdf (7.94 Ko) Télécharger le fichier
l2_recall_72_sift1m.pdf (8.08 Ko) Télécharger le fichier
recall@1_GIST1M.pdf (7.92 Ko) Télécharger le fichier
recall@1_SIFT1M.pdf (8.03 Ko) Télécharger le fichier
recall@1_VLAD500K.pdf (7.94 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Origin : Files produced by the author(s)
Origin : Files produced by the author(s)
Origin : Files produced by the author(s)
Origin : Files produced by the author(s)
Origin : Files produced by the author(s)
Origin : Files produced by the author(s)
Origin : Files produced by the author(s)
Origin : Files produced by the author(s)
Origin : Files produced by the author(s)
Origin : Files produced by the author(s)
Origin : Files produced by the author(s)
Origin : Files produced by the author(s)
Origin : Files produced by the author(s)
Origin : Files produced by the author(s)
Origin : Files produced by the author(s)
Origin : Files produced by the author(s)

Dates and versions

hal-01361953 , version 1 (07-09-2016)

Identifiers

  • HAL Id : hal-01361953 , version 1

Cite

Himalaya Jain, Patrick Pérez, Rémi Gribonval, Joaquin Zepeda, Hervé Jégou. Approximate search with quantized sparse representations. 14th European Conference on Computer Vision (ECCV), Oct 2016, Amsterdam, Netherlands. ⟨hal-01361953⟩
349 View
393 Download

Share

Gmail Facebook X LinkedIn More