Method Details
Details for method 'HRNetV2 + OCR'
Method overview
| name | HRNetV2 + OCR |
| challenge | pixel-level semantic labeling |
| details | HRNetV2W48 + OCR. OCR is an extension of object context networks https://arxiv.org/pdf/1809.00916.pdf |
| publication | High-Resolution Representations for Labeling Pixels and Regions; OCNet: Object Context Network for Scene Parsing HRNet Team; OCR Team https://arxiv.org/pdf/1904.04514.pdf |
| project page / code | https://github.com/HRNet |
| used Cityscapes data | fine annotations, coarse annotations |
| used external data | ImageNet, Mapillary |
| runtime | n/a |
| subsampling | no |
| submission date | June, 2019 |
| previous submissions |
Average results
| Metric | Value |
|---|---|
| IoU Classes | 83.3116 |
| iIoU Classes | 61.9549 |
| IoU Categories | 92.133 |
| iIoU Categories | 81.6543 |
Class results
| Class | IoU | iIoU |
|---|---|---|
| road | 98.8281 | - |
| sidewalk | 88.2199 | - |
| building | 94.2288 | - |
| wall | 67.6116 | - |
| fence | 65.323 | - |
| pole | 72.1673 | - |
| traffic light | 79.0762 | - |
| traffic sign | 82.3549 | - |
| vegetation | 94.1115 | - |
| terrain | 73.8429 | - |
| sky | 95.9778 | - |
| person | 88.1399 | 72.9969 |
| rider | 74.9823 | 54.0508 |
| car | 96.3672 | 91.2252 |
| truck | 76.8785 | 43.8217 |
| bus | 92.2904 | 59.0728 |
| train | 90.8524 | 57.816 |
| motorcycle | 72.8106 | 51.0184 |
| bicycle | 78.8577 | 65.6371 |
Category results
| Category | IoU | iIoU |
|---|---|---|
| flat | 98.774 | - |
| nature | 93.8351 | - |
| object | 77.6874 | - |
| sky | 95.9778 | - |
| construction | 94.374 | - |
| human | 88.2184 | 73.915 |
| vehicle | 96.0642 | 89.3936 |