Research map / Label-efficient and robust learning / Noisy labels
On Single Source Robustness in Deep Fusion Models
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Train AVOD with single-source noise and MaxSSN loss rather than clean-only training | AVOD with mean fusion · KITTI validation, Gaussian SSN, 3D car, easy | minAP3D (%): 47.41 -> 71.65 | Helped |
| Train AVOD with single-source noise and MaxSSN loss rather than clean-only training | AVOD with mean fusion · KITTI validation, Gaussian SSN, 3D car, moderate | minAP3D (%): 41.84 -> 62.14 | Helped |
| Use noise on all sources together (TrainASN) instead of single-source MaxSSN training (ablation) | AVOD with mean fusion · KITTI validation, Gaussian SSN, 3D car, easy | minAP3D (%): 71.65 -> 61.53 | Hurt |
| Replace mean fusion with the latent ensemble layer | AVOD · KITTI validation, Gaussian SSN, 3D car, easy | minAP3D (%): 47.41 -> 61.97 | Helped |
| Replace mean fusion with the latent ensemble layer | AVOD · KITTI validation, clean, 3D car, easy | AP3D (%): 76.41 -> 77.79 | Helped |
| Use single-source MaxSSN training rather than clean-only training | AVOD with LEL · KITTI validation, Gaussian SSN, 3D car, easy | minAP3D (%): 61.97 -> 68.16 | Helped |
| Use TrainSSN rather than TrainASN (ablation) | AVOD with LEL · KITTI validation, downsample SSN, 3D car, easy | minAP3D (%): 65.74 -> 73.33 | Helped |
| Use TrainSSN rather than TrainSSNAlt (ablation) | AVOD with LEL · KITTI validation, downsample SSN, 3D car, easy | minAP3D (%): 64.77 -> 73.33 | Helped |
| Train from scratch rather than fine-tune only fusion and prediction parameters (ablation) | AVOD with mean fusion · KITTI validation, Gaussian SSN, 3D car | minAP3D (%): 62.46 -> 71.65 | Helped |
| Train from scratch rather than fine-tune only fusion and prediction parameters (ablation) | AVOD with mean fusion · KITTI validation, Gaussian SSN, 3D car | minAP3D (%): 70.09 -> 71.66 | Helped |
| Use TrainSSN instead of TrainASN (ablation) | AVOD with mean fusion · KITTI validation, Gaussian SSN, 3D car, easy | maxDiffAP3D (%): 14.48 -> 3.71 | Helped |
Closest papers by meaning
- Fusion Is Not Enough: Single Modal Attacks on Fusion Models for 3D Object Detection (ICLR 2024 · similarity 0.85)
- COMMIT: Certifying Robustness of Multi-Sensor Fusion Systems Against Semantic Attacks (AAAI 2025 · similarity 0.85)
- Resilient Sensor Fusion Under Adverse Sensor Failures via Multi-Modal Expert Fusion (CVPR 2025 · similarity 0.83)
- Towards Trustworthy Deep Learning (AAAI 2024 · similarity 0.82)
- Adaptive Fusion of Single-View and Multi-View Depth for Autonomous Driving (CVPR 2024 · similarity 0.82)
- Cocoon: Robust Multi-Modal Perception with Uncertainty-Aware Sensor Fusion (ICLR 2025 · similarity 0.82)
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.