AI Paper Atlas

Research map / Detection and segmentation / Weakly supervised detection and segmentation

Self-supervised co-salient object detection via feature correspondences at multiple scales

Souradeep Chakraborty, Dimitris Samaras

ECCV 2024 · Detection and segmentation › Weakly supervised detection and segmentation · approach: Weakly supervised detection · 12 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
add region-level feature correspondence refinement (stage 2) (ablation)SCoSPARC · CoCAF-measure max: 0.567 -> 0.601Improved
add dense CRF post-processing (ablation)SCoSPARC · CoCAF-measure max: 0.601 -> 0.614Improved
add confidence-based adaptive thresholding instead of fixed 0.5 (ablation)SCoSPARC stage 1 · Cosal2015 / CoSOD3k / CoCAMAE and F-measure: Cosal2015 MAE 0.072 -> 0.069 but F 0.853 -> 0.840; CoCA F 0.564 -> 0.567Mixed
add saliency loss to co-occurrence loss (ablation)SCoSPARC stage 1 · CoCA vs Cosal2015/CoSOD3kF-measure max: CoCA 0.565 -> 0.564; CoSOD3k 0.801 -> 0.810Mixed
use patch size 16 instead of 8 in ViT-B (ablation)SCoSPARC stage 1 · CoCAF-measure max: 0.567 -> 0.511Hurt
use ViT-Small instead of ViT-Base (ablation)SCoSPARC stage 1 · CoCAF-measure max: 0.567 -> 0.560Hurt
use convolutional VGG-16 backbone instead of DINO ViT (ablation)SCoSPARC stage 1 · CoCAF-measure max: 0.567 -> 0.356Hurt
train on COCO9213 plus DUTS-Class instead of one dataset (ablation)SCoSPARC stage 1 · CoCAF-measure max: 0.532 -> 0.567 (vs COCO9213 only)Improved
self-supervised training without labels versus supervised GCoNet+ with 50% labelsSCoSPARC · CoCAF-measure max: 0.534 (GCoNet+ 50%) -> 0.614Improved
drop the dense CRF for speed (ablation)SCoSPARC · inference speedFPS: 4.1 -> 20.5 FPSImproved

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