Method Details


Details for method 'SERNet-Former_v2'

 

Method overview

name SERNet-Former_v2
challenge pixel-level semantic labeling
details Dense predictions for forward predictions are applied to attention mechanisms in SERNet-Former architecture, resulting in SERNet-Former_v2
publication Anonymous
project page / code
used Cityscapes data fine annotations, coarse annotations
used external data ImageNet
runtime n/a
subsampling no
submission date November, 2024
previous submissions 1, 2, 3, 4, 5, 6, 7

 

Average results

Metric Value
IoU Classes 85.0198
iIoU Classes 65.9104
IoU Categories 92.4796
iIoU Categories 81.7032

 

Class results

Class IoU iIoU
road 98.825 -
sidewalk 88.2768 -
building 94.6236 -
wall 72.6165 -
fence 69.5435 -
pole 73.2776 -
traffic light 78.7113 -
traffic sign 83.1717 -
vegetation 94.2232 -
terrain 74.6671 -
sky 96.1596 -
person 88.9162 73.5156
rider 76.6063 54.3099
car 96.533 90.574
truck 84.3302 52.7744
bus 95.3183 66.5975
train 92.7394 64.1954
motorcycle 77.0274 56.9322
bicycle 79.8099 68.384

 

Category results

Category IoU iIoU
flat 98.8176 -
nature 93.9852 -
object 78.5734 -
sky 96.1596 -
construction 94.6866 -
human 88.9239 74.241
vehicle 96.2107 89.1654

 

Links

Download results as .csv file

Benchmark page