Method Details
Details for method 'Deep Watershed Transformation'
Method overview
| name | Deep Watershed Transformation |
| challenge | instance-level semantic labeling |
| details | Instance segmentation using a watershed transformation inspired CNN. The input RGB image is augmented using the semantic segmentation from the recent PSPNet by H. Zhao et al. Previously named "DWT". |
| publication | Deep Watershed Transformation for Instance Segmentation Min Bai and Raquel Urtasun CVPR 2017 https://arxiv.org/abs/1611.08303 |
| project page / code | |
| used Cityscapes data | fine annotations |
| used external data | ImageNet |
| runtime | n/a |
| subsampling | 2 |
| submission date | April, 2017 |
| previous submissions | 1, 2 |
Average results
| Metric | Value |
|---|---|
| AP | 19.437 |
| AP50% | 35.3426 |
| AP100m | 31.4466 |
| AP50m | 36.8006 |
Class results
| Class | AP | AP50% | AP100m | AP50m |
|---|---|---|---|---|
| person | 15.5129 | 33.9822 | 27.6587 | 27.4405 |
| rider | 14.0919 | 36.8733 | 23.0767 | 23.7214 |
| car | 31.5441 | 48.5032 | 50.7829 | 53.4604 |
| truck | 22.5102 | 31.2823 | 37.9051 | 47.1182 |
| bus | 27.0348 | 40.057 | 46.3623 | 64.3084 |
| train | 22.9035 | 36.2295 | 33.6579 | 45.1376 |
| motorcycle | 13.9204 | 32.9197 | 19.3819 | 20.1667 |
| bicycle | 7.97816 | 22.8938 | 12.7473 | 13.0519 |