AI Paper Atlas

Research map / Label-efficient and robust learning / Knowledge distillation

Learning Lightweight Lane Detection CNNs by Self Attention Distillation

Yuenan Hou, Zheng Ma, Chunxiao Liu, Chen Change Loy

ICCV 2019 · Label-efficient and robust learning › Knowledge distillation · approach: Knowledge distillation for detectors · 826 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 SADENet · TuSimpleaccuracy: 93.02% -> 96.64%Improved
add SADResNet-18 · TuSimpleaccuracy: 92.69% -> 96.02%Improved
add SADResNet-34 · TuSimpleaccuracy: 92.84% -> 96.24%Improved
add SADResNet-101 · CULaneF1: 70.8 -> 71.8Improved
add SADResNet-101 · BDD100Kaccuracy: 34.45% -> 35.56%Improved
add SAD on low-level block 1 mimicking block 2 (ablation)ENet · TuSimpleaccuracy: 93.02% -> 91.22%Hurt
add SAD paths from middle and high level blocks (P23+P34) (ablation)ENet · TuSimpleaccuracy: 93.02% -> 96.64%Improved
mimic adjacent layer (P23+P34) rather than non-adjacent (P24+P34) (ablation)ENet-SAD · TuSimpleaccuracy: 96.52% -> 96.64%Improved
use backward distillation (higher layers mimic lower) (ablation)ENet · TuSimpleaccuracy: 93.02% -> 91.26%Hurt
use SAD instead of deep supervision (ablation)ENet · TuSimpleaccuracy: 94.69% -> 96.64%Improved
use SAD instead of deep supervision (ablation)ENet · BDD100KIoU: 15.38 -> 16.02Improved
block gradient of P1 branch to main branch (ablation)ENet-SAD · TuSimpleaccuracy: 96.64% -> 96.61%No clear effect

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