Method Details


Details for method 'FsaNet'

 

Method overview

name FsaNet
challenge pixel-level semantic labeling
details
publication FsaNet: Frequency Self-attention for Semantic Segmentation
project page / code
used Cityscapes data fine annotations, coarse annotations
used external data ImageNet
runtime n/a
subsampling no
submission date November, 2021
previous submissions

 

Average results

Metric Value
IoU Classes 83.0472
iIoU Classes 62.1924
IoU Categories 91.7694
iIoU Categories 79.5017

 

Class results

Class IoU iIoU
road 98.8011 -
sidewalk 87.8314 -
building 94.1029 -
wall 67.7092 -
fence 65.0912 -
pole 70.0636 -
traffic light 77.4609 -
traffic sign 80.7793 -
vegetation 93.9867 -
terrain 73.9698 -
sky 95.8864 -
person 87.9966 69.4674
rider 75.1001 53.4058
car 96.3836 90.8026
truck 79.2001 46.8414
bus 93.8519 60.3416
train 91.7521 63.0418
motorcycle 69.0274 48.1485
bicycle 78.9017 65.4904

 

Category results

Category IoU iIoU
flat 98.797 -
nature 93.7432 -
object 75.7854 -
sky 95.8864 -
construction 94.2116 -
human 87.9517 70.2421
vehicle 96.0106 88.7613

 

Links

Download results as .csv file

Benchmark page