AI Paper Atlas

Research map / Label-efficient and robust learning / Knowledge distillation

Pixel-Wise Contrastive Distillation

Junqiang Huang, Zichao Guo

ICCV 2023 · Label-efficient and robust learning › Knowledge distillation · approach: Knowledge distillation for detectors · 7 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 pixel-level instead of image-level contrastive distillation (ablation)PCD · COCO Mask R-CNN FPNAP bbox: 36.5 -> 37.4Improved
keep teacher projection head via SpatialAdaptor instead of removing it (ablation)PCD · VOC 07+12AP: 48.8 -> 52.1Improved
use channel-wise ReLU rather than ReLU in SpatialAdaptor (ablation)PCD · CityScapesmIoU: 70.1 -> 71.8Improved
add MHSA module to student (ablation)PCD · COCO Mask R-CNN FPNAP bbox: 37.1 -> 37.4Improved
replace MHSA with an extra prediction head of similar size (ablation)PCD · CityScapesmIoU: 71.8 -> 71.2Hurt
use symmetric loss on two views instead of asymmetric loss (ablation)PCD · COCO Mask R-CNN FPNAP bbox: 37.0 -> 37.4Improved
use SwAV, BYOL or Barlow Twins teacher instead of MoCo v3 (ablation)PCD · COCO Mask R-CNN FPNAP bbox: 37.4 -> 37.1No clear effect
use ViT-Base teacher (ablation)PCD · ImageNet linear probingtop-1 accuracy: 65.1 -> 57.6Hurt
use PCD instead of supervised pre-training for ResNet-50 studentResNet-50 · COCO Mask R-CNN FPNAP bbox: 40.2 -> 42.4Improved

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