Method Details


Details for method 'SegModel'

 

Method overview

name SegModel
challenge pixel-level semantic labeling
details FCN
publication Anonymous
project page / code
used Cityscapes data fine annotations
used external data ImageNet, MSCOCO
runtime 0.3 s
GTX TITAN X
subsampling no
submission date August, 2016
previous submissions

 

Average results

Metric Value
IoU Classes 75.4293
iIoU Classes 51.5043
IoU Categories 88.0324
iIoU Categories 72.7758

 

Class results

Class IoU iIoU
road 98.176 -
sidewalk 83.2037 -
building 91.7268 -
wall 51.0004 -
fence 53.9807 -
pole 52.6614 -
traffic light 66.2525 -
traffic sign 71.9588 -
vegetation 92.3155 -
terrain 69.0502 -
sky 94.7803 -
person 83.0612 59.6618
rider 66.8338 39.9865
car 94.8718 87.4722
truck 74.3119 35.4125
bus 79.3162 47.4778
train 72.2075 48.7032
motorcycle 65.663 40.0942
bicycle 71.7852 53.2263

 

Category results

Category IoU iIoU
flat 98.4262 -
nature 91.9579 -
object 61.8207 -
sky 94.7803 -
construction 92.0874 -
human 83.0609 60.387
vehicle 94.0934 85.1647

 

Links

Download results as .csv file

Benchmark page