Method Details


Details for method 'Seg_UJS'

 

Method overview

name Seg_UJS
challenge pixel-level semantic labeling
details Previously listed as "DaiLNet".
publication Anonymous
project page / code
used Cityscapes data fine annotations
used external data ImageNet
runtime n/a
subsampling no
submission date August, 2020
previous submissions 1, 2, 3

 

Average results

Metric Value
IoU Classes 82.6808
iIoU Classes 62.652
IoU Categories 92.2752
iIoU Categories 82.8834

 

Class results

Class IoU iIoU
road 98.7848 -
sidewalk 87.56 -
building 93.8803 -
wall 59.4408 -
fence 64.3149 -
pole 72.0563 -
traffic light 79.7484 -
traffic sign 82.6563 -
vegetation 94.1102 -
terrain 73.8658 -
sky 95.8797 -
person 88.7562 74.7341
rider 75.2358 54.4616
car 96.5989 91.4142
truck 76.081 44.9053
bus 90.3086 56.4768
train 90.7618 60.3073
motorcycle 71.8952 51.1707
bicycle 79.0003 67.7459

 

Category results

Category IoU iIoU
flat 98.8435 -
nature 93.8695 -
object 77.8117 -
sky 95.8797 -
construction 94.356 -
human 88.8938 75.857
vehicle 96.2723 89.9099

 

Links

Download results as .csv file

Benchmark page