AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Testing DNN-based Autonomous Driving Systems under Critical Environmental Conditions

Zhong Li, Minxue Pan, Tian Zhang, Xuandong Li

ICML 2021 · Image generation › Autoregressive and latent generation · approach: Generalizable fake image detection · 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 search-based (1+1) ES to find critical environmental conditions instead of random sampled conditions (Rc)TACTIC on Dave-orig · Udacity, five environmental typesKMNC and NBC coverage, number of errors: e.g., Night NBC 35.92% vs 3.18%; higher coverage in all typesImproved
Replace (1+1) ES search with random search (ablation)TACTIC on Dave-orig · Udacity, five environmental typesKMNC/NBC coverage and number of errors: Night NBC 10.73% vs 35.92%; errors at 40° 9.1 vs 4303.7Hurt
Use TACTIC (search-based critical conditions) instead of DeepRoadDave-orig · Udacity, Snow in DaytimeKMNC coverage: 54.41% vs 45.21% (TACTIC higher), but paper reports slightly lower coverage for TACTIC on Snow in DaytimeHurt
Use TACTIC instead of DeepRoadDave-orig · Udacity, five environmental typesnumber of erroneous behaviours at larger error bounds: Night 40° bound 4303.7 vs 9.0 for DeepRoadImproved
Use MUNIT-based TACTIC scenes instead of DeepTest and PreScan scenesimage realism · user study (34 participants)fraction of pairs where real scene judged more realistic: TACTIC 44% vs DeepTest 94%, PreScan 86%Improved

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