Method Details


Details for method 'pixel-to-anchor contrastive learning'

 

Method overview

name pixel-to-anchor contrastive learning
challenge pixel-level semantic labeling
details
publication Anonymous
project page / code
used Cityscapes data fine annotations
used external data
runtime n/a
subsampling no
submission date September, 2023
previous submissions

 

Average results

Metric Value
IoU Classes 79.61
iIoU Classes 59.8426
IoU Categories 91.4636
iIoU Categories 81.102

 

Class results

Class IoU iIoU
road 98.621 -
sidewalk 86.7307 -
building 93.3476 -
wall 54.8894 -
fence 61.605 -
pole 69.941 -
traffic light 77.7665 -
traffic sign 80.4566 -
vegetation 93.6213 -
terrain 73.6269 -
sky 95.4787 -
person 87.3306 71.9779
rider 73.6103 53.7101
car 96.2031 90.2008
truck 67.86 38.6696
bus 80.8337 56.9898
train 72.2456 50.4911
motorcycle 70.9065 50.7034
bicycle 77.5162 65.9982

 

Category results

Category IoU iIoU
flat 98.7682 -
nature 93.4179 -
object 75.7453 -
sky 95.4787 -
construction 93.6577 -
human 87.563 73.4044
vehicle 95.6143 88.7996

 

Links

Download results as .csv file

Benchmark page