Research map / Detection and segmentation / CNN features
RANK++LETR: Learn to Rank and Optimize Candidates for Line Segment Detection
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes RANK++LETR, an end-to-end line segment detector that learns to rank and then optimize line candidates to fix mismatch between confidence scores and localization quality.
Problem
In proposal-based line segment detection, confidence scores are predicted jointly with line locations, so they may not accurately reflect prediction quality.
Approach
Adds a learning-based line candidate ranking and optimization stage on top of proposal-based detection, so that candidates are re-ranked by predicted quality rather than relying on the simultaneously predicted confidence score.
Experimental setup
Builds on: LETR
Closest papers by meaning
- Improving Transformer Based Line Segment Detection with Matched Predicting and Re-ranking (AAAI 2025 · similarity 0.92)
- ScaleLSD: Scalable Deep Line Segment Detection Streamlined (CVPR 2025 · similarity 0.84)
- Polyline Path Masked Attention for Vision Transformer (NeurIPS 2025 · similarity 0.83)
- Breaking the Low-Rank Dilemma of Linear Attention (CVPR 2025 · similarity 0.82)
- DECO: Unleashing the Potential of ConvNets for Query-based Detection and Segmentation (ICLR 2025 · similarity 0.82)
- Learning Hierarchical Image Segmentation For Recognition and By Recognition (ICLR 2024 · similarity 0.82)
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.