Details for method 'Margin calibration'
Method overview
| name |
Margin calibration |
| challenge |
pixel-level semantic labeling |
| details |
The model is DeepLab v3+ backend on SEResNeXt50. We used the margin calibration with log-loss as the learning objective. |
| publication |
Anonymous
|
| project page / code |
|
| used Cityscapes data |
fine annotations, coarse annotations |
| used external data |
ImageNet |
| runtime |
n/a |
| subsampling |
no |
| submission date |
January, 2021 |
| previous submissions |
1 |
Average results
| Metric |
Value |
| IoU Classes | 82.0862 |
| iIoU Classes | 62.5345 |
| IoU Categories | 92.0798 |
| iIoU Categories | 81.7912 |
Class results
| Class |
IoU |
iIoU |
| road | 98.7328 | - |
| sidewalk | 87.3274 | - |
| building | 94.0389 | - |
| wall | 63.3662 | - |
| fence | 63.832 | - |
| pole | 71.6772 | - |
| traffic light | 78.1473 | - |
| traffic sign | 81.8323 | - |
| vegetation | 94.1323 | - |
| terrain | 73.5751 | - |
| sky | 96.0109 | - |
| person | 88.1464 | 72.8071 |
| rider | 72.8065 | 52.0566 |
| car | 96.525 | 91.6495 |
| truck | 74.9116 | 46.5095 |
| bus | 89.2873 | 61.5678 |
| train | 88.0946 | 60.5327 |
| motorcycle | 68.9006 | 49.795 |
| bicycle | 78.2938 | 65.3578 |
Category results
| Category |
IoU |
iIoU |
| flat | 98.8247 | - |
| nature | 93.8638 | - |
| object | 77.1818 | - |
| sky | 96.0109 | - |
| construction | 94.3402 | - |
| human | 88.212 | 73.5608 |
| vehicle | 96.1249 | 90.0215 |
Links
Download results as .csv file
Benchmark page
