Method Details


Details for method 'Knowledge-Aware'

 

Method overview

name Knowledge-Aware
challenge pixel-level semantic labeling
details Knowledge-Aware Semantic Segmentation
publication Anonymous
project page / code
used Cityscapes data fine annotations
used external data ImageNet
runtime n/a
subsampling no
submission date August, 2018
previous submissions

 

Average results

Metric Value
IoU Classes 79.2847
iIoU Classes 55.5564
IoU Categories 90.6937
iIoU Categories 78.0319

 

Class results

Class IoU iIoU
road 97.6299 -
sidewalk 80.0648 -
building 93.0373 -
wall 56.7747 -
fence 57.3128 -
pole 66.013 -
traffic light 72.5955 -
traffic sign 78.1643 -
vegetation 93.3898 -
terrain 71.0272 -
sky 95.6743 -
person 85.8842 67.6264
rider 69.6765 45.6335
car 95.9122 89.3389
truck 74.882 36.3195
bus 88.455 49.3772
train 86.3104 52.6141
motorcycle 68.144 43.6115
bicycle 75.4605 59.9304

 

Category results

Category IoU iIoU
flat 98.6906 -
nature 93.0977 -
object 72.573 -
sky 95.6743 -
construction 93.3881 -
human 85.9788 68.6779
vehicle 95.4536 87.386

 

Links

Download results as .csv file

Benchmark page