Method Details
Details for method 'Ladder DenseNet v2'
Method overview
| name | Ladder DenseNet v2 |
| challenge | pixel-level semantic labeling |
| details | DenseNet-121 model used in downsampling path with ladder-style skip connections upsampling path on top of it. |
| publication | Journal submission |
| project page / code | |
| used Cityscapes data | fine annotations |
| used external data | ImageNet |
| runtime | 1 s Titan X Pascal |
| subsampling | no |
| submission date | January, 2018 |
| previous submissions |
Average results
| Metric | Value |
|---|---|
| IoU Classes | 78.4016 |
| iIoU Classes | 54.572 |
| IoU Categories | 90.8488 |
| iIoU Categories | 78.7472 |
Class results
| Class | IoU | iIoU |
|---|---|---|
| road | 98.6595 | - |
| sidewalk | 86.6378 | - |
| building | 93.006 | - |
| wall | 52.9305 | - |
| fence | 56.9256 | - |
| pole | 67.1937 | - |
| traffic light | 74.4317 | - |
| traffic sign | 78.9016 | - |
| vegetation | 93.5731 | - |
| terrain | 73.1251 | - |
| sky | 95.4761 | - |
| person | 86.2371 | 68.2703 |
| rider | 69.6961 | 45.5005 |
| car | 95.9185 | 90.4978 |
| truck | 65.8715 | 35.4903 |
| bus | 82.1451 | 48.4828 |
| train | 76.3336 | 45.4023 |
| motorcycle | 66.8348 | 41.7123 |
| bicycle | 75.7337 | 61.2198 |
Category results
| Category | IoU | iIoU |
|---|---|---|
| flat | 98.7421 | - |
| nature | 93.2652 | - |
| object | 73.3422 | - |
| sky | 95.4761 | - |
| construction | 93.4379 | - |
| human | 86.3874 | 69.1188 |
| vehicle | 95.2907 | 88.3757 |