Method Details


Details for method 'C3Net [2,3,7,13]'

 

Method overview

name C3Net [2,3,7,13]
challenge pixel-level semantic labeling
details
publication C3: Concentrated-Comprehensive Convolution and its application to semantic segmentation
Hyojin Park, Youngjoon Yoo, Geonseok Seo, Dongyoon Han, Sangdoo Yun, Nojun Kwak
https://arxiv.org/abs/1812.04920
project page / code https://github.com/HYOJINPARK/C3SINet
used Cityscapes data
used external data
runtime n/a
subsampling 2
submission date September, 2019
previous submissions

 

Average results

Metric Value
IoU Classes 64.7831
iIoU Classes 37.3336
IoU Categories 84.6826
iIoU Categories 67.7562

 

Class results

Class IoU iIoU
road 97.2215 -
sidewalk 78.9608 -
building 88.6714 -
wall 40.8659 -
fence 43.5678 -
pole 52.8467 -
traffic light 54.7998 -
traffic sign 61.2993 -
vegetation 90.8207 -
terrain 67.5879 -
sky 93.5385 -
person 72.3015 52.0865
rider 50.8207 28.4382
car 90.739 83.827
truck 47.3121 18.2666
bus 56.1002 29.6186
train 43.5534 23.6793
motorcycle 41.9163 20.5648
bicycle 57.9556 42.1878

 

Category results

Category IoU iIoU
flat 97.8305 -
nature 90.4811 -
object 58.807 -
sky 93.5385 -
construction 88.8368 -
human 73.4743 53.8388
vehicle 89.8101 81.6736

 

Links

Download results as .csv file

Benchmark page