Method Details


Details for method 'SAITv2-light'

 

Method overview

name SAITv2-light
challenge pixel-level semantic labeling
details
publication Anonymous
project page / code
used Cityscapes data fine annotations, coarse annotations
used external data ImageNet + CITYSCAPES unlabeled data (except test-sequence)
runtime 0.025 s
GPU TITAN X
subsampling no
submission date June, 2018
previous submissions

 

Average results

Metric Value
IoU Classes 70.8749
iIoU Classes 41.5477
IoU Categories 86.6553
iIoU Categories 68.8548

 

Class results

Class IoU iIoU
road 98.1295 -
sidewalk 82.7927 -
building 90.7614 -
wall 49.3766 -
fence 50.2152 -
pole 54.3962 -
traffic light 59.0325 -
traffic sign 66.6689 -
vegetation 92.0208 -
terrain 70.5973 -
sky 94.4482 -
person 77.0459 53.9565
rider 57.0392 31.1237
car 93.6741 84.7518
truck 58.7016 23.8245
bus 73.905 35.8243
train 63.7349 32.461
motorcycle 49.3474 25.3335
bicycle 64.7349 45.1063

 

Category results

Category IoU iIoU
flat 98.3641 -
nature 91.7228 -
object 61.3553 -
sky 94.4482 -
construction 90.7909 -
human 77.3003 55.2592
vehicle 92.6054 82.4505

 

Links

Download results as .csv file

Benchmark page