Method Details


Details for method 'MCDN'

 

Method overview

name MCDN
challenge pixel-level semantic labeling
details
publication Anonymous
project page / code
used Cityscapes data fine annotations
used external data ImageNet
runtime n/a
subsampling no
submission date July, 2019
previous submissions

 

Average results

Metric Value
IoU Classes 81.1254
iIoU Classes 62.4173
IoU Categories 91.5161
iIoU Categories 81.2339

 

Class results

Class IoU iIoU
road 98.6827 -
sidewalk 87.0161 -
building 93.5427 -
wall 60.807 -
fence 65.6848 -
pole 69.8063 -
traffic light 77.132 -
traffic sign 80.8157 -
vegetation 93.8045 -
terrain 72.3652 -
sky 95.7801 -
person 87.0561 71.2233
rider 71.8961 53.7766
car 96.2117 91.5075
truck 77.5977 50.503
bus 88.7437 62.8994
train 77.4272 53.4329
motorcycle 70.3768 51.3029
bicycle 76.6357 64.6926

 

Category results

Category IoU iIoU
flat 98.7087 -
nature 93.4712 -
object 75.838 -
sky 95.7801 -
construction 93.9352 -
human 87.2327 72.4018
vehicle 95.6469 90.0661

 

Links

Download results as .csv file

Benchmark page