Research map / Detection and segmentation / Open-set and few-shot detection
Exploring the Sim2Real Gap Using Digital Twins
Sruthi Sudhakar, Jon Hanzelka, Josh Bobillot, Tanmay Randhavane, Neel Joshi, Vibhav Vineet
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 |
|---|---|---|---|
| reduce ambient light exposure of synthetic training data (ablation) | Faster-RCNN trained on YCB-Synthetic · YCB-Real | mAP: 81.93 -> 81.10 | No clear effect |
| add holes to object meshes in training data (ablation) | Faster-RCNN trained on YCB-Synthetic · YCB-Real | mAP: 81.93 -> 78.85 (holes 1) | Hurt |
| add holes to object meshes in training data (ablation) | Faster-RCNN trained on YCB-Synthetic · YCB-In-the-wild | mAP: 67.15 -> 55.19 (holes 3) | Hurt |
| add texture blur level 2-3 to training textures (ablation) | Faster-RCNN trained on YCB-Synthetic · YCB-Real | mAP: 81.93 -> 69.82 | Hurt |
| add mild texture blur (level 1) (ablation) | Faster-RCNN trained on YCB-Synthetic · YCB-Real | mAP: 81.93 -> 82.39 | No clear effect |
| use mismatched baked lighting in training renders (ablation) | Faster-RCNN trained on YCB-Synthetic · YCB-Real | mAP: 81.93 -> 72.87 | Hurt |
| add mesh noise level 3 (ablation) | Faster-RCNN trained on YCB-Synthetic · YCB-Real | mAP: 81.93 -> 65.65 | Hurt |
| add mesh noise level 3 (ablation) | Faster-RCNN trained on YCB-Synthetic · YCB-In-the-wild | mAP: 67.15 -> 20.30 | Hurt |
| add mesh noise level 3 (ablation) | Faster-RCNN trained on YCB-Synthetic · YCB-Video | mAP: 37.83 -> 18.88 | Hurt |
| reduce ambient light exposure of synthetic training data (ablation) | Faster-RCNN trained on YCB-Synthetic · YCB-Video | mAP: 37.83 -> 39.45 | No clear effect |
| evaluate on non-twin real data instead of the digital twin test set | Faster-RCNN trained on clean YCB-Synthetic · clean train set | mAP: 81.93 -> 67.15 -> 37.83 | Hurt |
Closest papers by meaning
- DAViD: Data-efficient and Accurate Vision Models from Synthetic Data (ICCV 2025 · similarity 0.89)
- AUGCAL: Improving Sim2Real Adaptation by Uncertainty Calibration on Augmented Synthetic Images (ICLR 2024 · similarity 0.86)
- Digital Twin Catalog: A Large-Scale Photorealistic 3D Object Digital Twin Dataset (CVPR 2025 · similarity 0.85)
- Real-Fake: Effective Training Data Synthesis Through Distribution Matching (ICLR 2024 · similarity 0.84)
- Scaling Laws of Synthetic Images for Model Training ... for Now (CVPR 2024 · similarity 0.84)
- Improving the Scaling Laws of Synthetic Data with Deliberate Practice (ICML 2025 · similarity 0.84)
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.
