Method Details
Details for method 'SGN'
Method overview
| name | SGN |
| challenge | instance-level semantic labeling |
| details | Instance segmentation using a sequence of neural networks, each solving a sub-grouping problem of increasing semantic complexity in order to gradually compose objects out of pixels. |
| publication | SGN: Sequential Grouping Networks for Instance Segmentation Shu Liu, Jiaya Jia, Sanja Fidler, Raquel Urtasun ICCV 2017 http://www.cs.toronto.edu/~urtasun/publications/liu_etal_iccv17.pdf |
| project page / code | |
| used Cityscapes data | fine annotations, coarse annotations |
| used external data | ImageNet |
| runtime | n/a |
| subsampling | no |
| submission date | March, 2017 |
| previous submissions |
Average results
| Metric | Value |
|---|---|
| AP | 25.0195 |
| AP50% | 44.91 |
| AP100m | 38.9448 |
| AP50m | 44.5019 |
Class results
| Class | AP | AP50% | AP100m | AP50m |
|---|---|---|---|---|
| person | 21.7638 | 45.2386 | 36.7275 | 36.8491 |
| rider | 20.0814 | 47.686 | 32.6583 | 33.268 |
| car | 39.4193 | 59.6584 | 60.0993 | 63.2262 |
| truck | 24.7639 | 36.3496 | 39.8556 | 50.7238 |
| bus | 33.2175 | 45.4391 | 53.7013 | 67.3556 |
| train | 30.818 | 53.6638 | 44.0998 | 59.2422 |
| motorcycle | 17.7346 | 39.4577 | 24.4113 | 25.3144 |
| bicycle | 12.3575 | 31.7869 | 20.0051 | 20.0359 |