Details for method 'HANet'
Method overview
| name |
HANet |
| challenge |
panoptic semantic labeling |
| details |
Holistic Attention Network for End-to-End Panoptic Segmentation |
| publication |
Anonymous
|
| project page / code |
|
| used Cityscapes data |
fine annotations |
| used external data |
ImageNet |
| runtime |
n/a |
| subsampling |
no |
| submission date |
May, 2019 |
| previous submissions |
|
Average results
| Metric |
All | Things | Stuff |
| PQ | 51.1519 | 40.4222 | 58.9552 |
| SQ | 77.6529 | 75.2629 | 79.3911 |
| RQ | 63.8859 | 53.5458 | 71.4059 |
Class results
| Class |
PQ | SQ | RQ |
| road | 97.555 | 97.7161 | 99.8351 |
| sidewalk | 69.8758 | 82.2592 | 84.9459 |
| building | 84.9481 | 88.3639 | 96.1345 |
| wall | 17.6264 | 69.6606 | 25.3033 |
| fence | 21.0319 | 69.4481 | 30.2844 |
| pole | 42.7458 | 65.3593 | 65.4013 |
| traffic light | 42.5336 | 69.0102 | 61.6338 |
| traffic sign | 60.7235 | 74.0722 | 81.9788 |
| vegetation | 89.0253 | 90.5018 | 98.3685 |
| terrain | 34.4927 | 74.7604 | 46.1377 |
| sky | 87.9492 | 92.15 | 95.4414 |
| person | 45.9815 | 73.8073 | 62.2994 |
| rider | 33.8722 | 64.3403 | 52.6454 |
| car | 57.1077 | 81.4004 | 70.1565 |
| truck | 39.4366 | 82.4583 | 47.8261 |
| bus | 46.8485 | 82.8215 | 56.5657 |
| train | 31.331 | 77.9258 | 40.2062 |
| motorcycle | 36.3249 | 71.6265 | 50.7143 |
| bicycle | 32.4755 | 67.7232 | 47.9532 |
Links
Download results as .csv file
Benchmark page
