Research map / Detection and segmentation / 3D detection and driving
Open-Weather Robust 3D Detection via Dual-Critic Diffusion Alignment
Shuyao Li, Chuanxing Geng, heyang sun, Qiang Zhou, Jingjing Gu
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add detection-guided critic to diffusion alignment (ablation) | DCDA · K-Radar type-open (unseen) | BEV AP (%): 66.23 -> 70.47 | Improved |
| add weather adversarial critic to diffusion alignment (ablation) | DCDA · K-Radar type-open (unseen) | BEV AP (%): 66.23 -> 68.22 | Improved |
| add weather adversarial critic on top of detection critic (ablation) | DCDA · K-Radar type-open (unseen) | BEV AP (%): 70.47 -> 72.65 | Improved |
| add detection critic on top of adversarial critic (ablation) | DCDA · K-Radar type-open (unseen) | 3D AP (%): 40.99 -> 42.25 | Improved |
| replace GAN refiner with diffusion aligner of same budget (ablation) | DCDA · K-Radar type-open (unseen) | BEV AP (%): 65.73 -> 72.65 | Improved |
| apply DCDA to every frame | LiDAR-4D radar detector · K-Radar type-open (unseen) | BEV AP (%): 65.96 -> 72.65 | Improved |
| add discriminator routing to bypass DCDA on Normal frames | DCDA · K-Radar type-open (unseen) | BEV AP (%): 72.65 -> 70.90 | Mixed |
| use DCDA instead of L4DR under held-out weather types | L4DR · K-Radar type-open (Fog) | BEV AP (%): 81.32 -> 88.81 | Improved |
| use DCDA instead of L4DR on unseen severity | L4DR · K-Radar severity-open | BEV/3D AP (%): +1.28 BEV AP, +2.32 3D AP | Improved |
| use DCDA instead of L4DR under type+severity holdout | L4DR · K-Radar type+severity-open | 3D AP (%): +2.18 mean unseen 3D AP | Improved |
| use DCDA instead of L4DR for real LightSnow to HeavySnow transfer | L4DR · K-Radar HeavySnow | BEV/3D AP (%): 30.57 -> 37.92 BEV AP, 16.57 -> 25.61 3D AP | Improved |
| use DCDA instead of L4DR in synthetic-to-real testing | L4DR · K-Radar real weather (severity-open trained) | BEV AP (%): +2.7 AP | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- L4DR — L4DR: LiDAR-4DRadar Fusion for Weather-Robust 3D Object Detection (AAAI 2025) · better on 12 of 12 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- An Empirical Study of the Generalization Ability of Lidar 3D Object Detectors to Unseen Domains (CVPR 2024 · similarity 0.92)
- L4DR: LiDAR-4DRadar Fusion for Weather-Robust 3D Object Detection (AAAI 2025 · similarity 0.90)
- Towards Robust 3D Object Detection with LiDAR and 4D Radar Fusion in Various Weather Conditions (CVPR 2024 · similarity 0.89)
- Sunshine to Rainstorm: Cross-Weather Knowledge Distillation for Robust 3D Object Detection (AAAI 2024 · similarity 0.89)
- WeatherGen: A Unified Diverse Weather Generator for LiDAR Point Clouds via Spider Mamba Diffusion (CVPR 2025 · similarity 0.88)
- UAWTrack: Universal 3D Single Object Tracking in Adverse Weather (AAAI 2025 · similarity 0.88)
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.
