AI Paper Atlas

Research map / Detection and segmentation / CNN features

Ultra Fast Structure-aware Deep Lane Detection

Zequn Qin, Huanyu Wang, Xi Li

ECCV 2020 · Detection and segmentation › CNN features · approach: Lane detection · 606 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
replace segmentation with row-based selecting formulation (ablation)Ours · TuSimple (module ablation, ResNet-34)accuracy: 92.84 -> 95.64Improved
add lane structural loss (ablation)Ours · TuSimple (module ablation, ResNet-34)accuracy: 95.64 -> 95.96Improved
add feature aggregation (ablation)Ours · TuSimple (module ablation, ResNet-34)accuracy: 95.64 -> 95.98Improved
add both structural loss and feature aggregation (ablation)Ours · TuSimple (module ablation, ResNet-34)accuracy: 95.64 -> 96.06Improved
use classification instead of regression (ablation)Ours · TuSimpleaccuracy: 71.59 -> 95.77Improved
normalize regression targets (ablation)Ours · TuSimpleaccuracy: 71.59 -> 67.24Hurt
use expectation-based localization instead of argmax (ablation)Ours · TuSimpleaccuracy: 95.77 -> 95.87Improved
increase number of gridding cells (ablation)Ours · TuSimpleclassification accuracyHurt
use lane similarity loss (ablation)Ours · TuSimpleaccuracyImproved
use row-based formulation vs plain segmentation with ResNet-34Res34 · TuSimple testruntime (ms): 50.5 -> 5.9Improved
use row-based formulation vs plain segmentation with ResNet-34Res34 · TuSimple testaccuracy: 92.84 -> 96.06Improved
use ResNet-34 instead of ResNet-18 backboneOurs · CULane test (total)F1-measure: 68.4 -> 72.3Improved

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