Add Toronto3D dataset

[Yarroudh]

I've trained KPFCNN model on Toronto3D Mobile LiDAR dataset using XYZRGB-Intensity. The results are presented as follow:

Overall Accuracy: 94.7% mean Intersection-over-Union: 79% IoU of each class (ignored unclassified):

1. Road: 96.6%
2. Road markings: 61.7%
3. Natural: 95.1%
4. Building: 89.1%
5. Utility line: 82.5%
6. Pole: 78.4%
7. Car: 87.1%
8. Fence: 41.8%

For comparison, the results reported by WeikaiTan using coordinates only:

OA: 95.39% mIoU: 69.11%

I can also provide the pretrained model if needed.