AI Paper Atlas

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

Beyond the Limitation of Monocular 3D Detector via Knowledge Distillation

Yiran Yang, Dongshuo Yin, Xuee Rong, Xian Sun, Wenhui Diao, Xinming Li

ICCV 2023 · Label-efficient and robust learning › Knowledge distillation · approach: Knowledge distillation for detectors · 7 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
use L2 instead of KL divergence for classification distillation (ablation)prediction distillation · KITTI val Car3D AP Mod: 11.18 -> 11.26No clear effect
add depth attention module to prediction distillation (ablation)L2 prediction distillation · KITTI val Car3D AP Mod: 11.26 -> 11.80Improved
weight objects by ground-truth depth with exponential mapping instead of linear (ablation)prediction distillation · KITTI val Car3D AP Mod: 11.45 -> 12.22Improved
add perspective-induced feature imitation to prediction distillation (ablation)depth-guided prediction distillation · KITTI val Car3D AP Mod: 12.25 -> 13.24Improved
add depth-guided prediction distillation to feature imitation (ablation)perspective-induced feature imitation · KITTI val Car3D AP Mod: 12.75 -> 13.24Improved
set perspective amplitude A=0.3 versus A=0.5 (ablation)perspective matrix · KITTI val CarBEV APMixed
apply distillation to student ResNet18PGD ResNet18 · KITTI val Car3D AP Mod: 11.61 -> 14.12Improved
apply distillation to MobileNetV2 student (T:ResNet101)PGD MobileNetV2 · KITTI val Car3D AP Mod: 7.09 -> 7.96Improved
apply distillation to SMOKE DLA34 studentSMOKE DLA34 · KITTI val Car3D AP Mod: 6.01 -> 7.79Improved
use depth-aware distillation versus FGD on nuScenesResNet18 student · nuScenes valmAP: 24.9 -> 25.1Improved

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