Method Details


Details for method 'DeepLab+DynamicCRF'

 

Method overview

name DeepLab+DynamicCRF
challenge pixel-level semantic labeling
details
publication ru.nl
project page / code
used Cityscapes data fine annotations
used external data ImageNet
runtime n/a
subsampling no
submission date July, 2016
previous submissions

 

Average results

Metric Value
IoU Classes 64.5466
iIoU Classes 38.3033
IoU Categories 83.6537
iIoU Categories 62.3921

 

Class results

Class IoU iIoU
road 96.836 -
sidewalk 77.8745 -
building 88.9061 -
wall 37.5468 -
fence 45.3701 -
pole 39.1124 -
traffic light 51.4838 -
traffic sign 61.5803 -
vegetation 90.7836 -
terrain 58.0039 -
sky 93.6473 -
person 76.6148 44.4664
rider 53.785 29.789
car 92.6038 81.3262
truck 41.8255 19.7566
bus 52.5075 31.2915
train 50.4994 30.4073
motorcycle 53.1818 28.6726
bicycle 64.2221 40.7169

 

Category results

Category IoU iIoU
flat 97.3453 -
nature 89.3509 -
object 48.154 -
sky 93.6473 -
construction 88.8186 -
human 77.1154 45.7756
vehicle 91.1442 79.0086

 

Links

Download results as .csv file

Benchmark page