Method Details


Details for method 'Hybrid Model'

 

Method overview

name Hybrid Model
challenge pixel-level semantic labeling
details
publication Anonymous
project page / code
used Cityscapes data fine annotations
used external data ImageNet
runtime n/a
subsampling no
submission date July, 2017
previous submissions

 

Average results

Metric Value
IoU Classes 65.7875
iIoU Classes 41.2319
IoU Categories 85.1868
iIoU Categories 68.5471

 

Class results

Class IoU iIoU
road 97.4695 -
sidewalk 78.4626 -
building 89.0354 -
wall 39.0286 -
fence 46.0631 -
pole 48.6452 -
traffic light 58.6987 -
traffic sign 64.0416 -
vegetation 91.2209 -
terrain 68.2504 -
sky 91.7694 -
person 76.8279 54.0806
rider 51.859 32.1671
car 92.2378 83.2834
truck 40.0273 22.0469
bus 50.5973 33.3811
train 44.882 24.3189
motorcycle 54.2714 31.2138
bicycle 66.5746 49.3632

 

Category results

Category IoU iIoU
flat 98.0677 -
nature 90.7687 -
object 56.849 -
sky 91.7694 -
construction 89.6314 -
human 78.128 55.5567
vehicle 91.0936 81.5376

 

Links

Download results as .csv file

Benchmark page