Depth Completion on TransCG (test)
0.012RMSETDCNet
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| TDCNetLearning Paradigm=Supervised Learning2025.12 | 0.012 | 0.017 | 0.008 | 92.37 | 97.98 | 99.81 | |
| TODE-TransLearning Paradigm=Supervised Learning2025.12 | 0.013 | 0.019 | 0.008 | 90.43 | 97.39 | 99.81 | |
| FDCTLearning Paradigm=Supervised Learning2025.12 | 0.015 | 0.022 | 0.01 | 88.18 | 97.15 | 99.81 | |
| DFNetGPU Memory Occupation=1.6 GB, Inference Time=0.0166s, Model Size=5.2 MB2022.02 | 0.018 | 0.027 | 0.012 | 0.8376 | 0.9567 | 0.9971 | |
| DFNetLearning Paradigm=Supervised Learning2025.12 | 0.018 | 0.027 | 0.012 | 83.76 | 95.67 | 99.71 | |
| LIDF-RefineGPU Memory Occupation=6.2 GB, Inference Time=0.0182s, Model Size=251 MB2022.02 | 0.019 | 0.034 | 0.015 | 0.7822 | 0.9426 | 0.998 | |
| TranspareNetGPU Memory Occupation=1.9 GB, Inference Time=0.0354s, Model Size=336 MB2022.02 | 0.026 | 0.023 | 0.013 | 0.8845 | 0.9625 | 0.9942 | |
| Ours (Self-supervised masking-based depth completion)Learning Paradigm=Fully Self-supervised Learning2025.12 | 0.026 | 0.029 | 0.014 | 81.74 | 94.43 | 99.8 | |
| MAELearning Paradigm=Fully Self-supervised Learning2025.12 | 0.029 | 0.037 | 0.018 | 73.88 | 92.29 | 99.79 | |
| DFNetTraining Dataset=Clear+OOD2022.02 | 0.039 | 0.067 | 0.03 | 56.68 | 74.61 | 98.01 | |
| ClearGraspGPU Memory Occupation=2.1 GB, Inference Time=2.2813s, Model Size=934 MB2022.02 | 0.054 | 0.083 | 0.037 | 0.5048 | 0.6868 | 0.9528 | |
| CGLearning Paradigm=Supervised Learning2025.12 | 0.054 | 0.083 | 0.037 | 50.48 | 68.68 | 95.28 | |
| [31]Training Dataset=Clear+OOD2022.02 | 0.061 | 0.108 | 0.049 | 33.59 | 54.73 | 92.48 | |
| [38]Training Dataset=Clear+OOD2022.02 | 0.071 | 0.06 | 0.036 | 62.99 | 82.92 | 95.93 | |
| LIDF-RefineTraining Dataset=Clear+OOD2022.02 | 0.146 | 0.262 | 0.115 | 13.7 | 26.39 | 57.95 |