AI Paper Atlas

Research map / Continual learning / Class-incremental

Class-Incremental Grouping Network for Continual Audio-Visual Learning

Shentong Mo, Weiguo Pian, Yapeng Tian

ICCV 2023 · Continual learning › Class-incremental · approach: Exemplar-efficient class-incremental learning · 27 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 audio-visual class tokens distillation (AVCTD) (ablation)CIGN baseline · VGGSound-100audio-visual Average Acc: 45.05 -> 51.02Improved
add audio-visual continual grouping (AVCG) only (ablation)CIGN baseline · VGGSound-100audio-visual Average Acc: 45.05 -> 48.72Improved
combine AVCTD and AVCG (ablation)CIGN · VGGSound-100audio-visual Average Acc / Forgetting: 45.05 -> 55.26 Acc; 15.26 -> 6.52 ForgettingImproved
increase transformer depth from 1 to 3 (ablation)AVCG · VGGSound-100audio-visual Average Acc: 51.25 -> 55.26Improved
increase transformer depth from 3 to 12 (ablation)AVCG · VGGSound-100audio-visual Average Acc: 55.26 -> 52.53Hurt
replace softmax with Gumbel hard-softmax in grouping (ablation)AVCG · VGGSound-100audio-visual Average Acc: 55.26 -> 48.95Hurt
increase buffer size per class from 10 to 50 (ablation)CIGN · VGGSound-100Average Acc and ForgettingImproved
use CIGN instead of DualPromptcontinual audio-visual learning · VGGSound-Instruments audio-visualAverage Acc: 63.32 -> 70.06Improved
use 10 incremental tasks instead of 4 (ablation)CIGN · VGGSound-100 audio-visualAverage Acc: 55.26 -> 48.16Hurt
apply CIGN with limited buffer on VGGSound-InstrumentsCIGN · VGGSound-Instrumentsgain over baselinesMixed

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