- The models are named as
model-dataset. - Model definition:
scale1denotes the 1/8 feature resolution model,scale2denotes the 1/8 & 1/4 model,scaleX-regrefineYdenotes theX-scale model with additionalYlocal regression refinements. - The inference time is averaged over 100 runs, measured with batch size 1 on a single NVIDIA A100 GPU.
- All pretrained models can be downloaded together at pretrained.zip, or they can be downloaded individually as listed below.
-
The inference time is measured for Sintel resolution: 448x1024
-
The
*-mixdatamodels are trained on several mixed public datasets, which are recommended for in-the-wild use cases.
| Model | Params (M) | Time (ms) | Download |
|---|---|---|---|
| GMFlow-scale1-things | 4.7 | 26 | download |
| GMFlow-scale1-mixdata | 4.7 | 26 | download |
| GMFlow-scale2-things | 4.7 | 66 | download |
| GMFlow-scale2-sintel | 4.7 | 66 | download |
| GMFlow-scale2-mixdata | 4.7 | 66 | download |
| GMFlow-scale2-regrefine6-things | 7.4 | 122 | download |
| GMFlow-scale2-regrefine6-sintelft | 7.4 | 122 | download |
| GMFlow-scale2-regrefine6-kitti | 7.4 | 122 | download |
| GMFlow-scale2-regrefine6-mixdata | 7.4 | 122 | download |
- The inference time is measured for KITTI resolution: 384x1248
- The
*-resumeflowthings-*denotes that the models are trained with GMFlow model as initialization, where GMFlow is trained on Chairs and Things dataset for optical flow task. - The
*-mixdatamodels are trained on several mixed public datasets, which are recommended for in-the-wild use cases.
| Model | Params (M) | Time (ms) | Download |
|---|---|---|---|
| GMStereo-scale1-sceneflow | 4.7 | 23 | download |
| GMStereo-scale1-resumeflowthings-sceneflow | 4.7 | 23 | download |
| GMStereo-scale2-sceneflow | 4.7 | 58 | download |
| GMStereo-scale2-resumeflowthings-sceneflow | 4.7 | 58 | download |
| GMStereo-scale2-regrefine3-sceneflow | 7.4 | 86 | download |
| GMStereo-scale2-regrefine3-resumeflowthings-sceneflow | 7.4 | 86 | download |
| GMStereo-scale2-regrefine3-resumeflowthings-kitti | 7.4 | 86 | download |
| GMStereo-scale2-regrefine3-resumeflowthings-middlebury | 7.4 | 86 | download |
| GMStereo-scale2-regrefine3-resumeflowthings-eth3dft | 7.4 | 86 | download |
| GMStereo-scale2-regrefine3-resumeflowthings-mixdata | 7.4 | 86 | download |
-
The inference time is measured for ScanNet resolution: 480x640
-
The
*-resumeflowthings-*models are trained with a pretrained GMFlow model as initialization, where GMFlow is trained on Chairs and Things dataset for optical flow task.
| Model | Params (M) | Time (ms) | Download |
|---|---|---|---|
| GMDepth-scale1-scannet | 4.7 | 17 | download |
| GMDepth-scale1-resumeflowthings-scannet | 4.7 | 17 | download |
| GMDepth-scale1-regrefine1-resumeflowthings-scannet | 4.7 | 20 | download |
| GMDepth-scale1-demon | 7.3 | 17 | download |
| GMDepth-scale1-resumeflowthings-demon | 7.3 | 17 | download |
| GMDepth-scale1-regrefine1-resumeflowthings-demon | 7.3 | 20 | download |