Method Details
Details for method 'kMaX-DeepLab [Cityscapes-fine]'
Method overview
| name | kMaX-DeepLab [Cityscapes-fine] |
| challenge | panoptic semantic labeling |
| details | kMaX-DeepLab w/ ConvNeXt-L backbone (ImageNet-22k + 1k pretrained). This result is obtained by the kMaX-DeepLab trained for Panoptic Segmentation task. No test-time augmentation or other external dataset. |
| publication | k-means Mask Transformer Qihang Yu, Huiyu Wang, Siyuan Qiao, Maxwell Collins, Yukun Zhu, Hartwig Adam, Alan Yuille, Liang-Chieh Chen ECCV 2022 https://arxiv.org/abs/2207.04044 |
| project page / code | https://github.com/google-research/deeplab2 |
| used Cityscapes data | fine annotations |
| used external data | ImageNet 22k + 1k |
| runtime | n/a |
| subsampling | no |
| submission date | March, 2022 |
| previous submissions |
Average results
| Metric | All | Things | Stuff |
|---|---|---|---|
| PQ | 66.165 | 59.5886 | 70.9479 |
| SQ | 84.0247 | 82.6127 | 85.0515 |
| RQ | 77.9242 | 71.9318 | 82.2824 |
Class results
| Class | PQ | SQ | RQ |
|---|---|---|---|
| road | 98.6459 | 98.711 | 99.9341 |
| sidewalk | 79.6408 | 86.3763 | 92.2022 |
| building | 89.8165 | 91.9254 | 97.7058 |
| wall | 45.5835 | 79.3626 | 57.4371 |
| fence | 45.864 | 78.4888 | 58.4337 |
| pole | 60.0506 | 73.5551 | 81.6403 |
| traffic light | 57.6235 | 78.6862 | 73.232 |
| traffic sign | 72.1203 | 82.6752 | 87.2332 |
| vegetation | 91.2221 | 92.1652 | 98.9768 |
| terrain | 48.6562 | 80.256 | 60.6262 |
| sky | 91.2031 | 93.3648 | 97.6847 |
| person | 56.0083 | 79.7248 | 70.2521 |
| rider | 56.6275 | 76.4088 | 74.1113 |
| car | 68.9085 | 86.201 | 79.9393 |
| truck | 57.658 | 88.4567 | 65.1822 |
| bus | 69.1579 | 89.3903 | 77.3663 |
| train | 67.7682 | 87.869 | 77.1242 |
| motorcycle | 53.414 | 78.6676 | 67.8984 |
| bicycle | 47.1663 | 74.1835 | 63.5805 |