Method Details


Details for method 'BatMAN_ROB'

 

Method overview

name BatMAN_ROB
challenge pixel-level semantic labeling
details batch-normalized multistage attention network
publication Anonymous
project page / code
used Cityscapes data fine annotations, coarse annotations
used external data ImageNet (pre-trained), KITTI, WildDash, ScanNet, Mapillary
runtime 1 s
GTX 1080, CPU
subsampling no
submission date May, 2018
previous submissions

 

Average results

Metric Value
IoU Classes 55.3763
iIoU Classes 29.2961
IoU Categories 83.8574
iIoU Categories 65.0231

 

Class results

Class IoU iIoU
road 97.4347 -
sidewalk 75.7454 -
building 88.1971 -
wall 29.0117 -
fence 34.7737 -
pole 48.3661 -
traffic light 49.3541 -
traffic sign 60.5688 -
vegetation 90.3257 -
terrain 63.3402 -
sky 94.229 -
person 66.286 44.4487
rider 16.8327 9.40593
car 91.196 83.5404
truck 20.9165 10.6296
bus 35.2045 24.7718
train 16.6145 8.16623
motorcycle 21.0684 13.6782
bicycle 52.6839 39.7281

 

Category results

Category IoU iIoU
flat 97.8712 -
nature 89.5256 -
object 55.0041 -
sky 94.229 -
construction 88.099 -
human 71.9824 48.1747
vehicle 90.2905 81.8714

 

Links

Download results as .csv file

Benchmark page