Mass estimation on ABO-500 (test)
7.53ADEPhysically Guided Visual Mass Estimation
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Physically Guided Visual Mass EstimationInput Modality=RGB+Depth, Fusion Method=self-attn.2026.01 | 7.53 | 1.019 | 0.9 | 0.458 | |
| Physically Guided Visual Mass EstimationInput Modality=RGB+Depth, Fusion Method=gated fusion2026.01 | 8.321 | 0.998 | 0.98 | 0.472 | |
| NeRF2PhysicsZero-shot=true2024.04 | 8.73 | 0.771 | 106.1 | 0.552 | |
| NeRF2PhysicsInput Modality=Multi-view input2026.01 | 8.73 | 0.771 | 1.061 | 0.552 | |
| Physically Guided Visual Mass EstimationInput Modality=RGB+Depth, Fusion Method=concat.2026.01 | 10.161 | 1.034 | 1.639 | 0.407 | |
| LLaVAInput Modality=RGB (VLM-based), Prediction Strategy=direct2026.01 | 12.083 | 3.157 | 2.312 | 0.28 | |
| Image2massStatus=Pretrained2024.04 | 12.496 | 1.792 | 97.6 | 0.341 | |
| image2massInput Modality=RGB2026.01 | 12.496 | 1.792 | 0.976 | 0.341 | |
| Cardoso and MorenoInput Modality=RGB+Depth2026.01 | 14.231 | 1.128 | 2.117 | 0.436 | |
| 2D CNN2024.04 | 15.431 | 1.609 | 1,445.9 | 0.362 | |
| RGB baselineInput Modality=RGB2026.01 | 15.431 | 1.609 | 14.459 | 0.362 | |
| LLaVA2024.04 | 17.328 | 1.893 | 183.7 | 0.306 | |
| Qwen-VLInput Modality=RGB (VLM-based), Prediction Strategy=direct2026.01 | 21.871 | 2.08 | 15.191 | 0.43 | |
| LLaVAInput Modality=RGB (VLM-based), Prediction Strategy=reasoning2026.01 | 30.588 | 1.781 | 12.283 | 0.368 | |
| Qwen-VLInput Modality=RGB (VLM-based), Prediction Strategy=reasoning2026.01 | 76.231 | 1.336 | 4.223 | 0.436 |