AI Paper Atlas

Research map / Continual learning / Online and replay-based

Anytime Continual Learning for Open Vocabulary Classification

Zhen Zhu, Yiming Gong, Derek Hoiem

ECCV 2024 · Continual learning › Online and replay-based · approach: Plug-and-play online training enhancements · 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 CLS-attention weighting to per-instance PCA (ablation)feature compression · CIFAR100FT accuracy: 74.6 -> 77.9Improved
use per-instance PCA instead of dataset-wide PCA (ablation)feature compression · CIFAR100FT accuracy and KB/example: 76.4 at 40.2 KB -> 77.9 at 19.4 KB (with CLS-weight)Improved
compress features with VQ instead of PCA (ablation)feature compression · CIFAR100FT accuracy: 77.8 -> 64.8Hurt
add int-quantization to PCA compression (ablation)feature compression · CIFAR100FT accuracy: 77.9 -> 77.5 and 19.4 -> 5.3 KB/exampleNo clear effect
use compressed features instead of full features for training (ablation)partial fine-tuning · CIFAR100ms/batch: 25.6 -> 13.9 ms/batch; 153.6 -> 5.3 KBImproved
replace AIM weighting with Online Class-wise Weighting (ablation)AnytimeCL · data and class incremental, flexible inferenceaccuracyImproved
tune only the last transformer block with fixed label embedding (ablation)AnytimeCL · flexible inferenceaccuracy on novel labelsImproved
class-balanced vs uniform sampling of stored samples (ablation)AnytimeCL · union data incrementalaccuracyNo clear effect
add the 'other' logit regularization (ablation)AnytimeCL · union data incrementalaccuracyImproved
use DINOv2 as the tuned model instead of CLIP (ablation)AnytimeCL · data incrementalaccuracyImproved
increase online batch size above 32 (ablation)AnytimeCL · union data incrementalaccuracyMixed
use AnytimeCL instead of LinProbe (AIM) on MTILtask incremental · MTIL Order IAvg.: 77.1 -> 77.0No clear effect

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