AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Deep Radial Embedding for Visual Sequence Learning

Yuecong Min, Peiqi Jiao, Yanan Li, Xiaotao Wang, Lei Lei, Xiujuan Chai, Xilin Chen

ECCV 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Contrastive learning for imbalanced data · 18 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 angle regularization to normalized CTC (ablation)CTC on Seq-MNIST · Seq-MNIST (balanced, dim=3)test accuracy: 94.8 -> 95.2Improved
add center regularization to normalized CTC (ablation)CTC on Seq-MNIST · Seq-MNIST (balanced, dim=3)test accuracy: 94.8 -> 95.6Improved
combine angle and center regularization (ablation)normalized CTC · Seq-MNIST (balanced, dim=3)test accuracy: 93.6Hurt
raise non-blank ratio eta from 0.0 to 0.8 (ablation)RadialCTC · Seq-MNIST (balanced, dim=3)localization mAP: 22.7 -> 87.9Improved
raise non-blank ratio eta from 0.0 to 0.8 (ablation)RadialCTC · Seq-MNIST (balanced, dim=3)test accuracy: 90.4 -> 80.8Hurt
ignore merged repeated labels at eta=0.8RadialCTC · Seq-MNIST (balanced, dim=3)test accuracy: 80.8 -> 94.6Improved
add full RadialCTC constraints on unbalanced data (ablation)CTC · Seq-MNIST (unbalanced, dim=3)test accuracy: 87.8 -> 73.6Hurt
use RadialCTC instead of CTC as intermediate supervisionBaseline [29] · Phoenix14 Test (multi-signer)WER: 22.6 -> 20.2Improved
use RadialCTC instead of CTC as intermediate supervisionBaseline [29] · Phoenix14 Test (signer-independent)WER: 33.8 -> 32.2Improved
replace CTC with RadialCTC as primary supervisionCRNN · IIIT5Kaccuracy: 79.8 -> 83.2Improved
replace CTC with RadialCTC as primary supervisionCRNN · SVTaccuracy: 80.4 -> 82.1Improved
replace CTC with RadialCTC as primary supervisionCRNN · IC13accuracy: 87.3 -> 90.7Improved

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