Details for method 'PGCNet_Res101_fine'
Method overview
| name |
PGCNet_Res101_fine |
| challenge |
pixel-level semantic labeling |
| details |
we choose the ResNet101 pretrained on ImageNet as our backbone, then we use both the train-fine and the val-fine data to train our model with batch size=8 for 8w iterations without any bells and whistles. We will release our paper latter. |
| publication |
Anonymous
|
| project page / code |
|
| used Cityscapes data |
fine annotations |
| used external data |
ImageNet |
| runtime |
n/a |
| subsampling |
no |
| submission date |
August, 2018 |
| previous submissions |
|
Average results
| Metric |
Value |
| IoU Classes | 80.5325 |
| iIoU Classes | 60.7475 |
| IoU Categories | 91.5117 |
| iIoU Categories | 81.1175 |
Class results
| Class |
IoU |
iIoU |
| road | 98.7467 | - |
| sidewalk | 87.149 | - |
| building | 93.6009 | - |
| wall | 59.5524 | - |
| fence | 62.6627 | - |
| pole | 69.251 | - |
| traffic light | 77.5977 | - |
| traffic sign | 80.3752 | - |
| vegetation | 93.9164 | - |
| terrain | 73.1299 | - |
| sky | 95.608 | - |
| person | 87.3425 | 72.164 |
| rider | 73.0713 | 53.8432 |
| car | 96.3583 | 90.5804 |
| truck | 70.7835 | 43.6926 |
| bus | 84.9056 | 57.7216 |
| train | 76.5046 | 49.4526 |
| motorcycle | 71.4723 | 52.0399 |
| bicycle | 78.0895 | 66.4859 |
Category results
| Category |
IoU |
iIoU |
| flat | 98.77 | - |
| nature | 93.5862 | - |
| object | 75.3184 | - |
| sky | 95.608 | - |
| construction | 93.9536 | - |
| human | 87.534 | 73.3423 |
| vehicle | 95.8116 | 88.8928 |
Links
Download results as .csv file
Benchmark page
