AI Paper Atlas

Research map / Continual learning / Class-incremental

Strike a Balance in Continual Panoptic Segmentation

Jinpeng Chen, Runmin Cong, Yuxuan Luo, Horace Ho Shing Ip, Sam Kwong

ECCV 2024 · Continual learning › Class-incremental · approach: Distillation-based continual segmentation · 5 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 past-class backtrace distillation (ablation)Mask2Former with pseudo-labels · ADE20K CPS 100-10PQ (all): 32.0 -> 34.7Improved
distill only past-class features instead of entire decoder features (ablation)BalConpas · ADE20K CPS 100-10PQ (all): 33.8 -> 34.7Improved
add entire-feature distillation vs no distillation (ablation)Mask2Former with pseudo-labels · ADE20K CPS 100-10PQ (all): 32.0 -> 33.8Improved
use class-proportional memory instead of random replay selection (ablation)BalConpas · ADE20K CPS 100-10PQ (all): 35.2 -> 36.5Improved
use class-proportional memory instead of SSUL equal-per-class selection (ablation)BalConpas · ADE20K CPS 100-10PQ (all): 35.8 -> 36.5Improved
add balanced anti-misguidance losses to replay (ablation)BalConpas with CPM · ADE20K CPS 100-10PQ (all): 35.1 -> 36.5Improved
base class distribution on segment counts instead of pixel counts (ablation)CPM · ADE20K CPS 100-10PQ (all): 35.2 -> 36.5Improved
base class distribution on segment counts instead of image counts (ablation)CPM · ADE20K CPS 100-10PQ (all): 36.0 -> 36.5Improved
add any replay selection strategy (random replay with BAG) (ablation)BalConpas without replay · ADE20K CPS 100-10PQ (all): 34.7 -> 35.2Improved
add replay with CPM and BAG to BalConpasBalConpas · ADE20K CPS 100-5PQ (all): 30.8 -> 32.6Improved
BalConpas versus EWF on average mIoUBalConpas · ADE20K CSS 100-5mIoU (avg): 43.4 -> 43.1Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

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