Method Details
Details for method 'ERFNet (from scratch)'
Method overview
| name | ERFNet (from scratch) |
| challenge | pixel-level semantic labeling |
| details | ERFNet trained from scratch only on the fine train (2975) annotated images |
| publication | Anonymous |
| project page / code | |
| used Cityscapes data | fine annotations |
| used external data | |
| runtime | 0.02 s 1 Titan X (Maxwell) |
| subsampling | 2 |
| submission date | January, 2017 |
| previous submissions |
Average results
| Metric | Value |
|---|---|
| IoU Classes | 67.3146 |
| iIoU Classes | 42.9913 |
| IoU Categories | 85.9597 |
| iIoU Categories | 71.7346 |
Class results
| Class | IoU | iIoU |
|---|---|---|
| road | 97.3908 | - |
| sidewalk | 79.6208 | - |
| building | 89.6563 | - |
| wall | 45.2995 | - |
| fence | 46.144 | - |
| pole | 55.1636 | - |
| traffic light | 61.2134 | - |
| traffic sign | 65.7921 | - |
| vegetation | 91.3026 | - |
| terrain | 68.5179 | - |
| sky | 93.5926 | - |
| person | 75.7618 | 58.541 |
| rider | 58.1249 | 33.9241 |
| car | 92.6665 | 85.0431 |
| truck | 46.0601 | 23.5801 |
| bus | 56.7429 | 35.4155 |
| train | 45.5143 | 29.5465 |
| motorcycle | 49.0455 | 30.1592 |
| bicycle | 61.3676 | 47.7211 |
Category results
| Category | IoU | iIoU |
|---|---|---|
| flat | 98.0269 | - |
| nature | 90.9303 | - |
| object | 61.741 | - |
| sky | 93.5926 | - |
| construction | 89.6734 | - |
| human | 76.4579 | 59.9839 |
| vehicle | 91.2955 | 83.4852 |