Details for method 'HRNetV2 + OCR (w/ ASP)'
Method overview
| name |
HRNetV2 + OCR (w/ ASP) |
| challenge |
pixel-level semantic labeling |
| details |
Our approach is based on a single HRNet48V2 and an OCR module combined with ASPP. We apply depth based multi-scale ensemble weights during testing (provided by DeepMotion AI Research) . |
| publication |
openseg-group (OCR team + HRNet team)
|
| project page / code |
https://github.com/openseg-group/openseg.pytorch
|
| used Cityscapes data |
fine annotations, coarse annotations |
| used external data |
ImageNet, Mapillary |
| runtime |
n/a |
| subsampling |
no |
| submission date |
July, 2019 |
| previous submissions |
|
Average results
| Metric |
Value |
| IoU Classes | 83.6704 |
| iIoU Classes | 64.8356 |
| IoU Categories | 92.3614 |
| iIoU Categories | 83.4671 |
Class results
| Class |
IoU |
iIoU |
| road | 98.829 | - |
| sidewalk | 88.2908 | - |
| building | 94.2629 | - |
| wall | 66.8827 | - |
| fence | 66.6902 | - |
| pole | 73.2846 | - |
| traffic light | 80.2195 | - |
| traffic sign | 83.0432 | - |
| vegetation | 94.2082 | - |
| terrain | 74.1028 | - |
| sky | 95.9733 | - |
| person | 88.5044 | 75.9988 |
| rider | 75.7896 | 57.4524 |
| car | 96.5108 | 91.6801 |
| truck | 78.5155 | 49.6362 |
| bus | 91.7864 | 62.0487 |
| train | 90.1252 | 58.3941 |
| motorcycle | 73.4026 | 55.283 |
| bicycle | 79.316 | 68.1914 |
Category results
| Category |
IoU |
iIoU |
| flat | 98.7911 | - |
| nature | 93.9219 | - |
| object | 78.6753 | - |
| sky | 95.9733 | - |
| construction | 94.4807 | - |
| human | 88.5539 | 76.8429 |
| vehicle | 96.1338 | 90.0913 |
Links
Download results as .csv file
Benchmark page
