Method Details


Details for method 'PartnerNet'

 

Method overview

name PartnerNet
challenge pixel-level semantic labeling
details PARTNERNET: A LIGHTWEIGHT AND EFFICIENT PARTNER NETWORK FOR SEMANTIC SEGMENTATION
publication Anonymous
project page / code
used Cityscapes data fine annotations
used external data
runtime 0.0058 s
GPU
subsampling no
submission date December, 2019
previous submissions

 

Average results

Metric Value
IoU Classes 72.3657
iIoU Classes 48.2787
IoU Categories 88.2401
iIoU Categories 73.5791

 

Class results

Class IoU iIoU
road 98.075 -
sidewalk 82.9759 -
building 91.171 -
wall 47.6685 -
fence 51.9677 -
pole 59.2035 -
traffic light 67.2042 -
traffic sign 70.383 -
vegetation 92.4592 -
terrain 70.2395 -
sky 94.6005 -
person 82.151 60.9546
rider 63.4685 40.0528
car 94.3749 86.7906
truck 53.3634 28.8193
bus 67.6938 42.5613
train 59.8161 36.1538
motorcycle 58.3008 34.8849
bicycle 69.831 56.0124

 

Category results

Category IoU iIoU
flat 98.4222 -
nature 92.1293 -
object 65.7374 -
sky 94.6005 -
construction 91.5377 -
human 82.0854 62.1709
vehicle 93.1684 84.9872

 

Links

Download results as .csv file

Benchmark page