Semantic Segmentation on KITTI-360 (val)
66.1mIoUImageTo360
Evaluation results
| Method | Links | |
|---|---|---|
| ImageTo360Backbone Networks=Segformer / SPVCNN, Use of images=yes, Annotation Amount=50%, Training Regime=Label Efficient2023.09 | 66.1 | |
| ImageTo360Backbone Networks=Segformer / Cylinder3D, Use of images=yes, Annotation Amount=50%, Training Regime=Label Efficient2023.09 | 65 | |
| ImageTo360Backbone Networks=Segformer / SPVCNN, Use of images=yes, Annotation Amount=20%, Training Regime=Label Efficient2023.09 | 64.2 | |
| SPTSize (x10^6)=0.777, superpoint-based=true2023.06 | 63.5 | |
| ImageTo360Backbone Networks=Segformer / SPVCNN, Use of images=yes, Annotation Amount=1%, Post-processing=TTA + Greedy Soup, Training Regime=Label Efficient2023.09 | 62.9 | |
| ImageTo360Backbone Networks=Segformer / SPVCNN, Use of images=yes, Annotation Amount=10%, Training Regime=Label Efficient2023.09 | 62.4 | |
| LaserMixBackbone Networks=Cylinder3D, Use of images=no, Annotation Amount=50%, Training Regime=Label Efficient2023.09 | 62.3 | |
| ImageTo360Backbone Networks=Segformer / Cylinder3D, Use of images=yes, Annotation Amount=20%, Training Regime=Label Efficient2023.09 | 62.2 | |
| DeepViewAggSize (x10^6)=41.22023.06 | 62.1 | |
| ImageTo360Backbone Networks=Segformer / SPVCNN, Use of images=yes, Annotation Amount=8%, Training Regime=Label Efficient2023.09 | 62.1 | |
| LaserMixBackbone Networks=Cylinder3D, Use of images=no, Annotation Amount=20%, Training Regime=Label Efficient2023.09 | 61.9 | |
| ImageTo360Backbone Networks=Segformer / SPVCNN, Use of images=yes, Annotation Amount=5%, Training Regime=Label Efficient2023.09 | 61.7 | |
| ScribbleKITTIBackbone Networks=SPVCNN, Use of images=no, Annotation Amount=8%, Training Regime=Weakly Annotated2023.09 | 61.3 | |
| ScribbleKITTIBackbone Networks=Cylinder3D, Use of images=no, Annotation Amount=8%, Training Regime=Weakly Annotated2023.09 | 60.8 | |
| LaserMixBackbone Networks=Cylinder3D, Use of images=no, Annotation Amount=10%, Training Regime=Label Efficient2023.09 | 60 | |
| ImageTo360Backbone Networks=Segformer / Cylinder3D, Use of images=yes, Annotation Amount=10%, Training Regime=Label Efficient2023.09 | 60 | |
| ReDALBackbone Networks=SPVCNN, Use of images=no, Annotation Amount=5%, Training Regime=Label Efficient2023.09 | 59.8 | |
| ImageTo360Backbone Networks=Segformer / Cylinder3D, Use of images=yes, Annotation Amount=8%, Training Regime=Label Efficient2023.09 | 59.6 | |
| LIDALBackbone Networks=SPVCNN, Use of images=no, Annotation Amount=5%, Training Regime=Label Efficient2023.09 | 59.5 | |
| ImageTo360Backbone Networks=Segformer / SPVCNN, Use of images=yes, Annotation Amount=1%, Training Regime=Label Efficient2023.09 | 59.5 | |
| ImageTo360Backbone Networks=Segformer / Cylinder3D, Use of images=yes, Annotation Amount=5%, Training Regime=Label Efficient2023.09 | 59.2 | |
| MinkowskiNetSize (x10^6)=37.92023.06 | 58.3 | |
| ImageTo360Backbone Networks=Segformer / SPVCNN, Use of images=yes, Annotation Amount=0.3%, Training Regime=Label Efficient2023.09 | 57.3 | |
| SPT-nanoSize (x10^6)=0.07, superpoint-based=true2023.06 | 57.2 | |
| ImageTo360Backbone Networks=Segformer / Cylinder3D, Use of images=yes, Annotation Amount=1%, Training Regime=Label Efficient2023.09 | 54.1 | |
| HybridCRBackbone Networks=PSD, Use of images=no, Annotation Amount=1%, Training Regime=Label Efficient2023.09 | 52.3 | |
| ImageTo360Backbone Networks=Segformer / SPVCNN, Use of images=yes, Annotation Amount=0%, Training Regime=Label Efficient2023.09 | 52.1 | |
| ImageTo360Backbone Networks=Segformer / Cylinder3D, Use of images=yes, Annotation Amount=0%, Training Regime=Label Efficient2023.09 | 50.8 | |
| ImageTo360Backbone Networks=Segformer / Cylinder3D, Use of images=yes, Annotation Amount=0.3%, Training Regime=Label Efficient2023.09 | 50.7 | |
| LaserMixBackbone Networks=Cylinder3D, Use of images=no, Annotation Amount=1%, Training Regime=Label Efficient2023.09 | 50.6 | |
| LIDALBackbone Networks=SPVCNN, Use of images=no, Annotation Amount=1%, Training Regime=Label Efficient2023.09 | 48.8 | |
| ST-SLidRBackbone Networks=SwAV / Minkowski, Use of images=yes, Annotation Amount=1%, Training Regime=Self-supervised Pretraining2023.09 | 44.9 | |
| SLidRBackbone Networks=Minkowski, Use of images=yes, Annotation Amount=1%, Training Regime=Self-supervised Pretraining2023.09 | 44.6 | |
| CLIP2SceneBackbone Networks=CLIP / SPVCNN, Use of images=yes, Annotation Amount=1%, Training Regime=Self-supervised Pretraining2023.09 | 42.6 | |
| Contrastive ImageBackbone Networks=ResNet18 / Minkowski, Use of images=yes, Annotation Amount=1%, Training Regime=Self-supervised Pretraining2023.09 | 42 | |
| ReDALBackbone Networks=SPVCNN, Use of images=no, Annotation Amount=1%, Training Regime=Label Efficient2023.09 | 41.8 |