AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

Self-Supervised Product Quantization for Deep Unsupervised Image Retrieval

Young Kyun Jang, Nam Ik Cho

ICCV 2021 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Invariance-based self-supervised learning · 83 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use cross quantized contrastive learning instead of comparing quantized outputs (ablation)SPQ · CIFAR-10 (32-bit)mAP: 0.763 -> 0.793Improved
use cross quantized contrastive learning instead of comparing quantized outputs (ablation)SPQ · FLICKR25K (32-bit)mAP: 0.751 -> 0.769Improved
use soft quantization instead of hard quantization (ablation)SPQ · CIFAR-10 (32-bit)mAP: 0.745 -> 0.793Improved
use soft quantization instead of hard quantization (ablation)SPQ · NUS-WIDE (32-bit)mAP: 0.742 -> 0.774Improved
use product quantization instead of standard vector quantization (ablation)SPQ · CIFAR-10 (32-bit)mAP: 0.734 -> 0.793Improved
use product quantization instead of standard vector quantization (ablation)SPQ · FLICKR25K (32-bit)mAP: 0.733 -> 0.769Improved
initialize with ImageNet pretrained weights (ablation)SPQ · CIFAR-10 (32-bit)mAP: 0.793 -> 0.814Improved
initialize with ImageNet pretrained weights (ablation)SPQ · NUS-WIDE (32-bit)mAP: 0.774 -> 0.788Improved
use ResNet backbone instead of VGG16 (ablation)SPQ · CIFAR-10 (32-bit)mAP: 0.761 -> 0.793Improved
use SPQ instead of semi-unsupervised TBHTBH · CIFAR-10 (32-bit)mAP: 0.573 -> 0.793Improved
use SPQ instead of BGANBGAN · FLICKR25K (32-bit)mAP: 0.686 -> 0.769Improved
vary tau_q and tau_cqc hyper-parameters (ablation)SPQ · CIFAR-10 (32-bit)mAPNo clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy