AI Paper Atlas

Research map / 3D vision / Depth and stereo

Improving Domain Generalization in Self-Supervised Monocular Depth Estimation via Stabilized Adversarial Training

Yuanqi Yao, Gang Wu, Kui Jiang, Siao Liu, Jian Kuai, Xianming Liu, Junjun Jiang

ECCV 2024 · 3D vision › Depth and stereo · approach: Self-supervised monocular depth · 4 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 SCAT (SDN + CGS adversarial training) to MonoDepth2MonoDepth2 · KITTI-CmCE (%): 101.04 -> 86.32Improved
add SCAT to MonoViTMonoViT · KITTI-CmCE (%): 80.54 -> 62.74Improved
add SCAT to Robust-DepthRobust-Depth · KITTI-CmCE (%): 55.72 -> 53.37Improved
use vanilla adversarial data augmentation instead of noneMonoViT · KITTI Eigen test (clean)Abs Rel: 0.099 -> 0.106Hurt
use SCAT rather than vanilla ADA on clean imagesMonoViT · KITTI Eigen test (clean)Abs Rel: 0.106 -> 0.100Improved
add SCAT to MonoDepth2 on night drivingMonoDepth2 · NuScenes-NightAbs Rel: 0.397 -> 0.355Improved
add offline data augmentation (DA) instead of adversarial augmentationMonoDepth2 · KITTI-CmCE (%): 101.04 -> 97.72Improved
add Conflict Gradient Surgery alone to adversarial training (ablation)MonoDepth2 ADA · KITTI-CAbs Rel: 0.193 -> 0.174Improved
add Scaling Depth Network alone to adversarial training (ablation)MonoDepth2 ADA · KITTI-CAbs Rel: 0.193 -> 0.179Improved
combine CGS and SDN (ablation)MonoDepth2 ADA · KITTI-CAbs Rel: 0.193 -> 0.165Improved
increase adversarial perturbation size beyond 135 (ablation)SCAT MonoDepth2 · KITTI-CAbs Rel: 0.165 -> 0.177Hurt
use small perturbation size 20 instead of 135 (ablation)SCAT MonoDepth2 · KITTI-CAbs Rel: 0.165 -> 0.196Hurt

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