Method Details


Details for method 'ARSAIT'

 

Method overview

name ARSAIT
challenge pixel-level semantic labeling
details anonymous
publication anonymous
anonymous
project page / code
used Cityscapes data fine annotations
used external data ImageNet
runtime 0.4 s
Titan Xp
subsampling no
submission date November, 2017
previous submissions

 

Average results

Metric Value
IoU Classes 62.2687
iIoU Classes 33.6048
IoU Categories 82.0076
iIoU Categories 61.8199

 

Class results

Class IoU iIoU
road 97.8675 -
sidewalk 81.1481 -
building 87.9926 -
wall 38.6582 -
fence 37.283 -
pole 35.5107 -
traffic light 48.8496 -
traffic sign 58.5156 -
vegetation 89.4181 -
terrain 68.6694 -
sky 93.0892 -
person 68.5054 43.6211
rider 44.1419 21.4012
car 91.4797 80.9909
truck 42.9282 18.0098
bus 51.5165 25.0973
train 48.1227 25.1413
motorcycle 41.1544 17.6641
bicycle 58.2549 36.9128

 

Category results

Category IoU iIoU
flat 97.9292 -
nature 89.1921 -
object 46.1758 -
sky 93.0892 -
construction 87.934 -
human 69.7692 45.3901
vehicle 89.9637 78.2497

 

Links

Download results as .csv file

Benchmark page