AI Paper Atlas

Research map / Detection and segmentation / CNN features

Adaptive Boundary Proposal Network for Arbitrary Shape Text Detection

Shi-Xue Zhang, Xiaobin Zhu, Chun Yang, Hongfa Wang, Xu-Cheng Yin

ICCV 2021 · Detection and segmentation › CNN features · approach: Arbitrary-shape scene text detection · 142 citations (OpenAlex)

Read the paper

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
use adaptive RNN+GCN encoder instead of RNN (ablation)deformation model · Total-TextF-measure: 85.93 -> 86.87Improved
use adaptive RNN+GCN encoder instead of graph convolution (ablation)deformation model · Total-TextF-measure: 86.19 -> 86.87Improved
use adaptive RNN+GCN encoder instead of FC (ablation)deformation model · CTW-1500F-measure: 81.54 -> 83.97Improved
use adaptive encoder instead of circular convolution (ablation)deformation model · CTW-1500F-measure: 82.55 -> 83.97Improved
use adaptive encoder instead of RNN (ablation)deformation model · CTW-1500F-measure: 83.56 -> 83.97Improved
add distance field to classification map (ablation)prior information · Total-TextF-measure: 79.87 -> 85.32Improved
add direction field to classification and distance (ablation)prior information · Total-TextF-measure: 85.32 -> 86.87Improved
increase inference iterations from 1 to 3 (ablation)boundary deformation · CTW-1500F-measure: 82.24 -> 83.97Improved
increase inference iterations from 1 to 3 (ablation)boundary deformation · CTW-1500FPS: 13.68 -> 12.08Hurt
increase iterations from 2 to 3 (ablation)boundary deformation · CTW-1500F-measure: 83.33 -> 83.97No clear effect
use higher-resolution FPN features (1/1 vs 1/4) (ablation)shared features · Total-TextF-measure: 86.13 -> 86.87Improved
use higher-resolution FPN features (1/1 vs 1/4) (ablation)shared features · Total-TextFPS: 15.17 -> 10.56Hurt

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