Details for method 'AttaNet_light'
Method overview
| name |
AttaNet_light |
| challenge |
pixel-level semantic labeling |
| details |
|
| publication |
AttaNet: Attention-Augmented Network for Fast and Accurate Scene Parsing(AAAI21)
|
| project page / code |
https://github.com/songqi-github/AttaNet
|
| used Cityscapes data |
fine annotations |
| used external data |
|
| runtime |
n/a |
| subsampling |
no |
| submission date |
August, 2020 |
| previous submissions |
|
Average results
| Metric |
Value |
| IoU Classes | 70.0869 |
| iIoU Classes | 42.925 |
| IoU Categories | 87.0132 |
| iIoU Categories | 72.1335 |
Class results
| Class |
IoU |
iIoU |
| road | 97.8086 | - |
| sidewalk | 80.8466 | - |
| building | 90.482 | - |
| wall | 49.5574 | - |
| fence | 49.6204 | - |
| pole | 51.719 | - |
| traffic light | 62.7009 | - |
| traffic sign | 67.575 | - |
| vegetation | 92.1429 | - |
| terrain | 69.3504 | - |
| sky | 93.9575 | - |
| person | 80.8009 | 58.0372 |
| rider | 60.9699 | 33.474 |
| car | 93.9577 | 88.1617 |
| truck | 61.657 | 22.5167 |
| bus | 62.8188 | 34.5175 |
| train | 47.003 | 27.1936 |
| motorcycle | 51.067 | 28.4279 |
| bicycle | 67.6161 | 51.0717 |
Category results
| Category |
IoU |
iIoU |
| flat | 98.424 | - |
| nature | 91.8718 | - |
| object | 59.6544 | - |
| sky | 93.9575 | - |
| construction | 90.998 | - |
| human | 81.0943 | 58.8215 |
| vehicle | 93.0922 | 85.4454 |
Links
Download results as .csv file
Benchmark page
