Details for method 'FoveaNet'
Method overview
| name |
FoveaNet |
| challenge |
pixel-level semantic labeling |
| details |
I trained a fully convolutional neural network with ResNet-101 as the front-end model. |
| publication |
Anonymous
|
| project page / code |
|
| used Cityscapes data |
fine annotations |
| used external data |
ImageNet |
| runtime |
n/a |
| subsampling |
no |
| submission date |
September, 2016 |
| previous submissions |
1, 2 |
Average results
| Metric |
Value |
| IoU Classes | 73.0693 |
| iIoU Classes | 49.1746 |
| IoU Categories | 88.6032 |
| iIoU Categories | 73.7766 |
Class results
| Class |
IoU |
iIoU |
| road | 98.1288 | - |
| sidewalk | 82.8166 | - |
| building | 91.2361 | - |
| wall | 45.2125 | - |
| fence | 49.8478 | - |
| pole | 60.602 | - |
| traffic light | 67.2351 | - |
| traffic sign | 73.5028 | - |
| vegetation | 92.4569 | - |
| terrain | 69.813 | - |
| sky | 94.6962 | - |
| person | 81.9123 | 61.1638 |
| rider | 60.6944 | 40.8631 |
| car | 94.3419 | 86.8431 |
| truck | 58.6355 | 30.8458 |
| bus | 68.3464 | 40.8869 |
| train | 66.273 | 40.5358 |
| motorcycle | 60.9204 | 36.573 |
| bicycle | 71.6444 | 55.6854 |
Category results
| Category |
IoU |
iIoU |
| flat | 98.4428 | - |
| nature | 92.1713 | - |
| object | 67.4526 | - |
| sky | 94.6962 | - |
| construction | 91.5945 | - |
| human | 82.9032 | 62.7408 |
| vehicle | 92.9621 | 84.8124 |
Links
Download results as .csv file
Benchmark page
