Released pre-trained models for all main paper experiments.
Classes | Number of meanshapes |
Masks | Paper reference |
3D IoU |
Download |
---|---|---|---|---|---|
Aeroplane | 8 | GT + MaskRCNN | Table 2 | 0.460 | model |
Car | 10 | GT + MaskRCNN | Table 2 | 0.684 | model |
Aeroplane,Car | 2 | GT + MaskRCNN | Table 2 | 0.567 | model |
Aeroplane,Car | 1 | GT + PointRend | Table 4 | 0.532 | model |
Aeroplane,Car | 2 | GT + PointRend | Table 4 | 0.552 | model |
Bicycle,Bus,Car,Motorbike | 1 | GT + PointRend | Table 4 | 0.517 | model |
Bicycle,Bus,Car,Motorbike | 4 | GT + PointRend | Table 4 | 0.543 | model |
12 Pascal3D+ classes | 1 | GT + PointRend | Table 4 | 0.409 | model |
12 Pascal3D+ classes | 12 | GT + PointRend | Table 4 | 0.425 | model |
To compute the 3D IoU metric for Pascal3D+, click here and download the MatLab code. We started from the DRC repo and adapt the code for reading our generated mat files containing the predicted meshes.
Classes | Number of meanshapes |
Masks | Paper reference |
Mask IoU |
Download |
---|---|---|---|---|---|
Bird | 1 | GT | Table 3 | 0.658 | model |
Bird | 14 | GT | Table 3 | 0.642 | model |