Method Details
Details for method 'Ladder DenseNet'
Method overview
| name | Ladder DenseNet |
| challenge | pixel-level semantic labeling |
| details | https://ivankreso.github.io/publication/ladder-densenet/ |
| publication | Ladder-style DenseNets for Semantic Segmentation of Large Natural Images Ivan Krešo, Josip Krapac, Siniša Šegvić ICCV 2017 https://ivankreso.github.io/publication/ladder-densenet/ |
| project page / code | https://github.com/ivankreso/ladder-densenet |
| used Cityscapes data | fine annotations |
| used external data | ImageNet |
| runtime | 0.45 s 2 x GTX 1070 |
| subsampling | no |
| submission date | March, 2017 |
| previous submissions |
Average results
| Metric | Value |
|---|---|
| IoU Classes | 74.3022 |
| iIoU Classes | 51.5884 |
| IoU Categories | 89.6593 |
| iIoU Categories | 79.4787 |
Class results
| Class | IoU | iIoU |
|---|---|---|
| road | 97.4339 | - |
| sidewalk | 80.2016 | - |
| building | 91.9555 | - |
| wall | 47.5906 | - |
| fence | 53.9237 | - |
| pole | 64.5682 | - |
| traffic light | 72.8064 | - |
| traffic sign | 76.2656 | - |
| vegetation | 92.8061 | - |
| terrain | 66.4327 | - |
| sky | 95.5204 | - |
| person | 83.8312 | 68.8379 |
| rider | 66.1288 | 42.9057 |
| car | 94.3409 | 90.1004 |
| truck | 55.605 | 28.9906 |
| bus | 70.2602 | 42.4892 |
| train | 66.9691 | 40.7319 |
| motorcycle | 62.0758 | 38.2156 |
| bicycle | 73.0267 | 60.4361 |
Category results
| Category | IoU | iIoU |
|---|---|---|
| flat | 98.263 | - |
| nature | 92.1204 | - |
| object | 71.1285 | - |
| sky | 95.5204 | - |
| construction | 92.2585 | - |
| human | 84.4627 | 70.3593 |
| vehicle | 93.8615 | 88.598 |