Models of semantic segmentation applied in KITTI road dataset. I use SD-Filter for LiDAR upsampling.
| Loss | Train Loss | Test Loss |
|---|---|---|
| **** |  |
 |
| Result | Origin Images | Predict Images |
|---|---|---|
| **** |  |
 |
| Model | FCNxs | Late-fusion FCN(minus) | cross-fusion FCN(cat) | Cross-fusion FCN(2 layers' cross-plus) | cross-fusion FCN(all layers' cross-plus) | PSPNet |
|---|---|---|---|---|---|---|
| MaxF | 95.05 | 94.94 | 95.78 | 95.58 | 93.79 | 97.97 |
| AvgPrec | 89.95 | 90.36 | 90.06 | 88.93 | 88.24 | 90.60 |