Method Details


Details for method 'SAIT'

 

Method overview

name SAIT
challenge pixel-level semantic labeling
details Trained on full resolution images. Anonymous submission.
publication Anonymous
project page / code
used Cityscapes data fine annotations
used external data ImageNet
runtime 3 s
GPU TITAN X
subsampling no
submission date April, 2016
previous submissions

 

Average results

Metric Value
IoU Classes 66.8218
iIoU Classes 39.3494
IoU Categories 86.0865
iIoU Categories 67.8489

 

Class results

Class IoU iIoU
road 97.7372 -
sidewalk 80.1947 -
building 89.7554 -
wall 36.3931 -
fence 42.4334 -
pole 50.558 -
traffic light 60.2682 -
traffic sign 66.3154 -
vegetation 91.1434 -
terrain 69.4056 -
sky 94.4327 -
person 77.7923 51.6488
rider 54.8784 31.2122
car 92.7039 84.5916
truck 45.1177 20.9409
bus 57.195 31.5235
train 42.8271 21.9479
motorcycle 53.5678 26.9866
bicycle 66.8941 45.9438

 

Category results

Category IoU iIoU
flat 98.2308 -
nature 90.8727 -
object 58.6374 -
sky 94.4327 -
construction 89.91 -
human 78.8474 53.3062
vehicle 91.6743 82.3916

 

Links

Download results as .csv file

Benchmark page