Details for method 'NJUST'
Method overview
| name |
NJUST |
| challenge |
instance-level semantic labeling |
| details |
Mask R-CNN based on FPN enhancement and Mask Rescore, etc. Only one single model SE-ResNext-152 with COCO pre-train used; |
| publication |
Ang Li, Chongyang Zhang
|
| project page / code |
|
| used Cityscapes data |
fine annotations |
| used external data |
ImageNet, COCO |
| runtime |
n/a |
| subsampling |
no |
| submission date |
March, 2019 |
| previous submissions |
|
Average results
| Metric |
Value |
| AP | 38.9387 |
| AP50% | 64.1239 |
| AP100m | 53.0113 |
| AP50m | 55.4106 |
Class results
| Class |
AP | AP50% | AP100m | AP50m |
| person | 44.0281 | 75.9711 | 60.8682 | 60.8815 |
| rider | 35.2486 | 70.4722 | 49.3374 | 50.0185 |
| car | 57.8658 | 81.0638 | 75.846 | 77.5723 |
| truck | 36.1813 | 48.1047 | 48.3495 | 53.2271 |
| bus | 48.7033 | 65.5302 | 68.3771 | 75.92 |
| train | 35.0994 | 57.8341 | 47.8267 | 52.0235 |
| motorcycle | 30.4988 | 58.361 | 39.7527 | 40.0378 |
| bicycle | 23.884 | 55.6545 | 33.7324 | 33.6038 |
Links
Download results as .csv file
Benchmark page
