Method Details
Details for method 'NVSegNet'
Method overview
| name | NVSegNet |
| challenge | pixel-level semantic labeling |
| details | In the inference, we use the image of 2 different scales. The same for training! |
| publication | Anonymous |
| project page / code | |
| used Cityscapes data | fine annotations |
| used external data | ImageNet |
| runtime | 0.4 s 2 GPU (1 titan and 1 quadro) |
| subsampling | no |
| submission date | May, 2016 |
| previous submissions | 1, 2 |
Average results
| Metric | Value |
|---|---|
| IoU Classes | 67.4338 |
| iIoU Classes | 41.3607 |
| IoU Categories | 87.2378 |
| iIoU Categories | 68.0985 |
Class results
| Class | IoU | iIoU |
|---|---|---|
| road | 97.9908 | - |
| sidewalk | 81.8509 | - |
| building | 90.1094 | - |
| wall | 35.682 | - |
| fence | 39.8044 | - |
| pole | 57.4041 | - |
| traffic light | 60.6322 | - |
| traffic sign | 69.2825 | - |
| vegetation | 91.7433 | - |
| terrain | 67.6186 | - |
| sky | 94.5529 | - |
| person | 79.2562 | 51.864 |
| rider | 54.4874 | 33.2894 |
| car | 93.5097 | 84.9556 |
| truck | 43.7789 | 25.5545 |
| bus | 52.4167 | 34.54 |
| train | 50.3039 | 25.3112 |
| motorcycle | 52.9967 | 27.7776 |
| bicycle | 67.8209 | 47.5933 |
Category results
| Category | IoU | iIoU |
|---|---|---|
| flat | 98.3876 | - |
| nature | 91.5879 | - |
| object | 63.5407 | - |
| sky | 94.5529 | - |
| construction | 90.4621 | - |
| human | 80.1654 | 53.5429 |
| vehicle | 91.9678 | 82.6541 |