Method Details
Details for method 'ERFNet (from scratch)'
Method overview
| name | ERFNet (from scratch) |
| challenge | pixel-level semantic labeling |
| details | ERFNet trained entirely on the fine train set (2975 images) without any pretraining nor coarse labels |
| publication | Efficient ConvNet for Real-time Semantic Segmentation Eduardo Romera, Jose M. Alvarez, Luis M. Bergasa and Roberto Arroyo IV2017 http://www.robesafe.uah.es/personal/eduardo.romera/pdfs/Romera17iv.pdf |
| project page / code | https://github.com/Eromera/erfnet |
| used Cityscapes data | fine annotations |
| used external data | |
| runtime | 0.02 s Titan X (Maxwell) |
| subsampling | 2 |
| submission date | February, 2017 |
| previous submissions | 1 |
Average results
| Metric | Value |
|---|---|
| IoU Classes | 68.0173 |
| iIoU Classes | 40.4237 |
| IoU Categories | 86.4645 |
| iIoU Categories | 70.4017 |
Class results
| Class | IoU | iIoU |
|---|---|---|
| road | 97.7366 | - |
| sidewalk | 80.9941 | - |
| building | 89.8331 | - |
| wall | 42.4633 | - |
| fence | 47.9916 | - |
| pole | 56.245 | - |
| traffic light | 59.836 | - |
| traffic sign | 65.2822 | - |
| vegetation | 91.3833 | - |
| terrain | 68.1993 | - |
| sky | 94.1887 | - |
| person | 76.7534 | 56.6562 |
| rider | 57.075 | 31.4706 |
| car | 92.7603 | 84.9471 |
| truck | 50.7691 | 19.3994 |
| bus | 60.0891 | 35.0554 |
| train | 51.8043 | 24.9675 |
| motorcycle | 47.2739 | 24.266 |
| bicycle | 61.6499 | 46.6273 |
Category results
| Category | IoU | iIoU |
|---|---|---|
| flat | 98.1759 | - |
| nature | 91.1173 | - |
| object | 62.4208 | - |
| sky | 94.1887 | - |
| construction | 90.0565 | - |
| human | 77.4251 | 58.0242 |
| vehicle | 91.8675 | 82.7791 |