Details for method 'EdgeSenseSeg'
Method overview
| name |
EdgeSenseSeg |
| challenge |
pixel-level semantic labeling |
| details |
Deep segmentation network with hard negative mining and other tricks. |
| publication |
Anonymous
|
| project page / code |
|
| used Cityscapes data |
fine annotations |
| used external data |
|
| runtime |
n/a |
| subsampling |
no |
| submission date |
October, 2017 |
| previous submissions |
|
Average results
| Metric |
Value |
| IoU Classes | 76.7648 |
| iIoU Classes | 57.1459 |
| IoU Categories | 89.7833 |
| iIoU Categories | 78.4579 |
Class results
| Class |
IoU |
iIoU |
| road | 98.3857 | - |
| sidewalk | 84.7894 | - |
| building | 92.454 | - |
| wall | 51.9914 | - |
| fence | 58.1482 | - |
| pole | 61.483 | - |
| traffic light | 72.9611 | - |
| traffic sign | 76.0664 | - |
| vegetation | 93.2508 | - |
| terrain | 71.8155 | - |
| sky | 95.0447 | - |
| person | 85.2392 | 68.3589 |
| rider | 68.5316 | 49.119 |
| car | 95.3796 | 88.8405 |
| truck | 62.3966 | 41.7955 |
| bus | 77.5506 | 53.1845 |
| train | 70.6995 | 46.861 |
| motorcycle | 66.8293 | 46.672 |
| bicycle | 75.5153 | 62.3355 |
Category results
| Category |
IoU |
iIoU |
| flat | 98.5929 | - |
| nature | 92.8865 | - |
| object | 68.9191 | - |
| sky | 95.0447 | - |
| construction | 92.9248 | - |
| human | 85.5948 | 69.6591 |
| vehicle | 94.5205 | 87.2567 |
Links
Download results as .csv file
Benchmark page
