AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

ZoomNeXt: A Unified Collaborative Pyramid Network for Camouflaged Object Detection

Youwei Pang, Xiaoqi Zhao, Tian-Zhu Xiang, Lihe Zhang, Huchuan Lu

TPAMI 2024 · Detection and segmentation › Open-set and few-shot detection · approach: Open-set and unknown object detection · 148 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 RGPU to baseline (ablation)ZoomNeXt baseline · COD10KS-measure: 0.826 -> 0.843 (G=6)Improved
add MHSIU (M=4) to baseline (ablation)ZoomNeXt baseline · COD10KS-measure: 0.826 -> 0.856Improved
add both RGPU and MHSIU (ablation)ZoomNeXt baseline · COD10KS-measure: 0.826 -> 0.861; ARG 6.56%Improved
add uncertainty awareness loss (UAL) (ablation)ZoomNeXt (RGPU + MHSIU) · COD10Kweighted F-measure: 0.734 -> 0.768Improved
increase RGPU groups from 6 to 8 (ablation)RGPU · COD10KS-measure: 0.843 -> 0.839Hurt
replace RGPU with Res2Net block (ablation)ZoomNeXt baseline · COD10KS-measure: 0.843 -> 0.833Hurt
increase MHSIU heads from 4 to 8 (ablation)MHSIU · COD10Kweighted F-measure: 0.728 -> 0.726No clear effect
use mixed scales 0.5x,1.0x,1.5x with MHSIU instead of single 1.0x input (ablation)ZoomNeXt baseline · COD10KS-measure: 0.826 -> 0.856Improved
use MHSIU instead of addition for fusing three scales (ablation)mixed-scale input · COD10Kweighted F-measure: 0.719 -> 0.728Improved
replace max-pooling down-sampling with bi-linear interpolation (ablation)MHSIU · COD10KS-measure: 0.856 -> 0.852; ARG down 2.32%Hurt
use cosine instead of constant lambda schedule for UAL (ablation)UAL · COD10KS-measure: 0.859 -> 0.861No clear effect
increase number of input frames from T=1 to T=5 (ablation)ZoomNeXt · MoCA-Mask-TES-measure: 0.690 -> 0.734Improved

Later papers that compared with this method

7 later accepted papers report ZoomNeXt as a baseline in their results tables; 7 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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