Method Details


Details for method 'MultiBoost'

 

Method overview

name MultiBoost
challenge pixel-level semantic labeling
details Boosting based solution. Publication is under review.
publication Anonymous
project page / code
used Cityscapes data fine annotations, coarse annotations, stereo
used external data ImageNet
runtime 0.25 s
GPU: Nvidia 980 Ti
subsampling 2
submission date January, 2017
previous submissions

 

Average results

Metric Value
IoU Classes 59.2623
iIoU Classes 32.4639
IoU Categories 81.8763
iIoU Categories 60.2343

 

Class results

Class IoU iIoU
road 95.8988 -
sidewalk 69.461 -
building 87.2636 -
wall 34.414 -
fence 32.7105 -
pole 40.4997 -
traffic light 54.8839 -
traffic sign 58.6386 -
vegetation 89.2169 -
terrain 65.2762 -
sky 90.2933 -
person 68.4305 41.2403
rider 42.5322 25.7827
car 89.0131 77.8001
truck 22.5494 11.2095
bus 51.8556 23.5953
train 40.858 24.9419
motorcycle 36.5065 21.7529
bicycle 55.6822 33.3881

 

Category results

Category IoU iIoU
flat 97.4555 -
nature 88.7325 -
object 50.9119 -
sky 90.2933 -
construction 87.0524 -
human 71.3531 44.9924
vehicle 87.3354 75.4762

 

Links

Download results as .csv file

Benchmark page