Method Details


Details for method 'ThunderNet'

 

Method overview

name ThunderNet
challenge pixel-level semantic labeling
details
publication Anonymous
project page / code
used Cityscapes data fine annotations
used external data ImageNet
runtime 0.0104 s
Titan XP
subsampling 2
submission date March, 2018
previous submissions

 

Average results

Metric Value
IoU Classes 63.9713
iIoU Classes 40.4482
IoU Categories 84.1166
iIoU Categories 69.3088

 

Class results

Class IoU iIoU
road 97.2891 -
sidewalk 77.3601 -
building 88.2901 -
wall 41.1941 -
fence 38.3957 -
pole 48.5266 -
traffic light 55.6007 -
traffic sign 60.8346 -
vegetation 90.6684 -
terrain 67.7284 -
sky 92.9753 -
person 71.5587 53.6992
rider 46.5816 30.2889
car 91.5822 83.9644
truck 39.3194 21.0494
bus 49.9156 34.936
train 49.8387 32.565
motorcycle 45.5089 26.5057
bicycle 62.2871 40.5768

 

Category results

Category IoU iIoU
flat 97.8975 -
nature 90.2785 -
object 56.0441 -
sky 92.9753 -
construction 88.3876 -
human 73.4009 56.0402
vehicle 89.8321 82.5773

 

Links

Download results as .csv file

Benchmark page