Method Details
Details for method 'FCN 8s'
Method overview
| name | FCN 8s |
| challenge | pixel-level semantic labeling |
| details | Trained by Marius Cordts on a pre-release version of the dataset |
| publication | Fully Convolutional Networks for Semantic Segmentation J. Long, E. Shelhamer, and T. Darrell CVPR 2015 http://www.cv-foundation.org/openaccess/content_cvpr_2015/papers/Long_Fully_Convolutional_Networks_2015_CVPR_paper.pdf |
| project page / code | http://fcn.berkeleyvision.org/ |
| used Cityscapes data | fine annotations |
| used external data | ImageNet, Pascal Context |
| runtime | 0.5 s Nvidia Titan X |
| subsampling | no |
| submission date | March, 2016 |
| previous submissions |
Average results
| Metric | Value |
|---|---|
| IoU Classes | 65.3254 |
| iIoU Classes | 41.7038 |
| IoU Categories | 85.6742 |
| iIoU Categories | 70.1393 |
Class results
| Class | IoU | iIoU |
|---|---|---|
| road | 97.406 | - |
| sidewalk | 78.4065 | - |
| building | 89.2114 | - |
| wall | 34.9328 | - |
| fence | 44.2369 | - |
| pole | 47.4143 | - |
| traffic light | 60.0832 | - |
| traffic sign | 65.0173 | - |
| vegetation | 91.4171 | - |
| terrain | 69.2969 | - |
| sky | 93.8604 | - |
| person | 77.1373 | 55.9347 |
| rider | 51.4129 | 33.3743 |
| car | 92.628 | 83.9126 |
| truck | 35.2722 | 22.2475 |
| bus | 48.5751 | 30.7803 |
| train | 46.5414 | 26.6649 |
| motorcycle | 51.569 | 31.0826 |
| bicycle | 66.7635 | 49.6335 |
Category results
| Category | IoU | iIoU |
|---|---|---|
| flat | 98.2462 | - |
| nature | 91.1283 | - |
| object | 57.0125 | - |
| sky | 93.8604 | - |
| construction | 89.6176 | - |
| human | 78.5849 | 57.9571 |
| vehicle | 91.2693 | 82.3215 |