Method Details


Details for method 'Mask R-CNN [fine-only]'

 

Method overview

name Mask R-CNN [fine-only]
challenge instance-level semantic labeling
details Mask R-CNN, ResNet-50-FPN, Cityscapes fine-only
publication Mask R-CNN
Kaiming He, Georgia Gkioxari, Piotr Dollár, Ross Girshick
https://arxiv.org/abs/1703.06870
project page / code
used Cityscapes data fine annotations
used external data ImageNet
runtime n/a
subsampling no
submission date April, 2017
previous submissions

 

Average results

Metric Value
AP 26.2171
AP50% 49.8889
AP100m 37.6317
AP50m 40.1117

 

Class results

Class AP AP50% AP100m AP50m
person 30.516 60.7121 46.2497 46.2318
rider 23.748 59.546 35.5611 35.8849
car 46.8876 68.3012 65.4649 67.3948
truck 22.7572 33.1146 31.0879 37.7669
bus 32.1504 48.2429 45.9532 51.1989
train 18.615 38.8608 27.5327 32.8783
motorcycle 19.1004 46.4623 24.8629 25.2702
bicycle 15.9618 43.8713 24.3413 24.2675

 

Links

Download results as .csv file

Benchmark page