AI Paper Atlas

Research map / Adversarial robustness and security / Attacks on text and graphs

Adversarial Training of Self-supervised Monocular Depth Estimation against Physical-World Attacks

Zhiyuan Cheng, James Chenhao Liang, Guanhong Tao, Dongfang Liu, Xiangyu Zhang

ICLR 2023 · Adversarial robustness and security › Attacks on text and graphs · approach: Latency and driving-perception attacks · 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 L0 adversarial training with self-supervised view-synthesis reconstructionMonodepth2 · L0 1/10 attackABSE: 6.08 -> 0.25Improved
Use L0 adversarial training with self-supervised view-synthesis reconstructionDepthHints · L0 1/10 attackABSE: 3.19 -> 0.27Improved
Use L0 perturbation training rather than no hardeningMonodepth2 · PGD 0.1 attackABSE: 11.68 -> 1.53Improved
Use L0 adversarial training with self-supervised view-synthesis reconstructionMonodepth2 · Patch attackABSE: 2.71 -> 0.39Improved
Use L0 adversarial training with self-supervised view-synthesis reconstructionMonodepth2 · KITTI Eigen splitABSE: 2.125 -> 2.16No clear effect
Replace self-supervised reconstruction training with contrastive learningMonodepth2 · KITTI Eigen splitABSE: 2.125 -> 3.218Hurt
Train with distance range 5-10 meters rather than other tested ranges (ablation)L0+SelfSup Monodepth2 · Adversarial distance range 5-30 metersABSEImproved
Randomize training viewing angles over -30° to 30° rather than fix the angle at 0° (ablation)L0+SelfSup Monodepth2 · L0 1/10 attackABSE: 0.271 -> 0.251Improved
Expand the randomized training angle range from -30°–30° to -40°–40° (ablation)L0+SelfSup Monodepth2 · L0 1/10 attackABSE: 0.251 -> 0.249Improved
Train from scratch instead of fine-tuning (ablation)L0+SelfSup Monodepth2 · Adversarial attacksABSEImproved
Train from scratch instead of fine-tuning (ablation)L0+SelfSup Monodepth2 · Benign depth estimationABSE: 2.16 -> 2.468Hurt
Add contrastive learning to self-supervised adversarial training (ablation)Monodepth2 · L0 1/10 attackABSE: 0.25 -> 0.244Improved

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