FKAConv
About
| Category | Paris-Lille-3D |
| Short name | FKAConv |
| Long name | FKAConv: Feature-Kernel Alignment Convolution |
| URL | |
| Description | FKAConv: Feature-Kernel Alignment for Point Cloud Convolution -- https://arxiv.org/abs/2004.04462 |
| BibTeX | @article{boulch2020fkaconv, title={FKAConv: Feature-Kernel Alignment for Point Cloud Convolution}, author={Boulch, Alexandre and Puy, Gilles and Marlet, Renaud}, journal={arXiv preprint arXiv:2004.04462}, year={2020} } |
| Hardware | RTX2080Ti |
| Runtime | 6000 |
| Used additional training data | Yes |
| Source code | No |
| Created at | 11/11/2019 19:55 |
| Updated at | 08/07/2020 07:55 |
Result
| Av. IoU | Ground | Building | Pole | Bollard | Trash_can | Barrier | Pedestrian | Car | Natural | |
|---|---|---|---|---|---|---|---|---|---|---|
| Result | 82.7 | 99.6 | 98.1 | 77.2 | 91.1 | 64.7 | 66.5 | 58.1 | 95.6 | 93.9 |
| Rank | 4 | 2 | 1 | 2 | 1 | 7 | 3 | 10 | 3 | 3 |
Full confusion matrix
| L1' | L2' | L3' | L4' | L5' | L6' | L7' | L8' | L9' | |
|---|---|---|---|---|---|---|---|---|---|
| L1 | 18095073 | 4907 | 307 | 113 | 33 | 4058 | 13 | 1148 | 12428 |
| L2 | 29735 | 6688409 | 2 | 1 | 2 | 1626 | 0 | 0 | 21487 |
| L3 | 816 | 10917 | 122334 | 198 | 0 | 2082 | 0 | 1003 | 18649 |
| L4 | 262 | 0 | 0 | 14263 | 76 | 253 | 0 | 0 | 61 |
| L5 | 432 | 8 | 0 | 140 | 25569 | 4370 | 0 | 3338 | 3 |
| L6 | 15784 | 7957 | 1381 | 275 | 5096 | 229942 | 41 | 20977 | 28693 |
| L7 | 461 | 4647 | 3 | 8 | 16 | 79 | 15452 | 70 | 5818 |
| L8 | 2323 | 1240 | 22 | 0 | 312 | 1345 | 2 | 933147 | 11244 |
| L9 | 4847 | 49388 | 690 | 0 | 139 | 21953 | 6 | 385 | 2704236 |