AI Paper Atlas

Research map / 3D vision / 3D generation

SurfelGAN: Synthesizing Realistic Sensor Data for Autonomous Driving

Zhenpei Yang, Yuning Chai, Dragomir Anguelov, Yin Zhou, Pei Sun, Dumitru Erhan, Sean Rafferty, Henrik Kretzschmar

CVPR 2020 · 3D vision › 3D generation · approach: 3D perception for autonomous driving · 107 citations (OpenAlex)

Check 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 direct surfel renderings with SurfelGAN-SAC imagesVehicle detector evaluated on WOD-EVAL · WOD-EVALAP@50: 0.521 -> 0.620Improved
Add adversarial loss to paired supervised training (ablation)SurfelGAN-S · WOD-EVALAP@50: 0.576 -> 0.610Improved
Add unpaired data and cycle consistency to supervised adversarial training (ablation)SurfelGAN-SA · WOD-EVALAP@50: 0.610 -> 0.620Improved
Add unpaired data and cycle consistency to supervised adversarial training (ablation)SurfelGAN-SA · DCPL1-distance: 0.240 -> 0.238Hurt
Add adversarial loss to paired supervised training (ablation)SurfelGAN-S · DCPL1-distance: 0.229 -> 0.240Hurt
Add WOD-TRAIN-NV surfel renderings to detector trainingVehicle detector · WOD-EVALAP@50: 0.219 -> 0.228Improved
Add WOD-TRAIN-NV SurfelGAN images to detector trainingVehicle detector · WOD-EVALAP@50: 0.219 -> 0.254Improved
Increase novel-view pose perturbationSurfelGAN-SAC · WOD-EVAL-NV, d <= 1.0 versus 2.0 < dAP@50: 0.574 -> 0.488Hurt
Use surfel inputs with coverage ratio at or below 0.3 rather than coverage above 0.5SurfelGAN · WOD-EVAL-NVAP@50: 0.566 -> 0.490Hurt
Use SurfelGAN-S instead of direct surfel renderings (ablation)Image synthesis evaluated on DCP · DCPL1-distance: 0.262 -> 0.229Improved
Add cycle consistency and unpaired training data (ablation)SurfelGAN-SA · WOD-EVAL-NVAP@50: 0.567 -> 0.570Improved

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