Method Details
Details for method 'EMANet'
Method overview
| name | EMANet |
| challenge | pixel-level semantic labeling |
| details | |
| publication | Expectation Maximization Attention Networks for Semantic Segmentation Xia Li, Zhisheng Zhong, Jianlong Wu, Yibo Yang, Zhouchen Lin, Hong Liu ICCV 2019 https://arxiv.org/abs/1907.13426 |
| project page / code | |
| used Cityscapes data | fine annotations |
| used external data | ImageNet |
| runtime | n/a |
| subsampling | no |
| submission date | December, 2019 |
| previous submissions |
Average results
| Metric | Value |
|---|---|
| IoU Classes | 81.9255 |
| iIoU Classes | 61.3399 |
| IoU Categories | 91.6306 |
| iIoU Categories | 80.3402 |
Class results
| Class | IoU | iIoU |
|---|---|---|
| road | 98.7106 | - |
| sidewalk | 87.2993 | - |
| building | 93.8247 | - |
| wall | 63.4468 | - |
| fence | 62.2666 | - |
| pole | 70.0468 | - |
| traffic light | 77.9459 | - |
| traffic sign | 80.7243 | - |
| vegetation | 93.9258 | - |
| terrain | 73.6308 | - |
| sky | 95.6938 | - |
| person | 87.8477 | 71.3495 |
| rider | 74.4901 | 53.622 |
| car | 96.2093 | 90.4202 |
| truck | 75.5303 | 44.5477 |
| bus | 90.2383 | 59.5953 |
| train | 84.4824 | 54.0041 |
| motorcycle | 71.5467 | 51.0686 |
| bicycle | 78.7241 | 66.1121 |
Category results
| Category | IoU | iIoU |
|---|---|---|
| flat | 98.7106 | - |
| nature | 93.5913 | - |
| object | 75.9106 | - |
| sky | 95.6938 | - |
| construction | 94.0303 | - |
| human | 87.8334 | 72.1344 |
| vehicle | 95.6444 | 88.546 |