Method Details


Details for method 'FoveaNet'

 

Method overview

name FoveaNet
challenge pixel-level semantic labeling
details 1.caffe-master 2.resnet-101 3.single scale testing Previously listed as "LXFCRN".
publication FoveaNet
Xin Li, Jiashi Feng
project page / code
used Cityscapes data fine annotations
used external data ImageNet
runtime n/a
subsampling no
submission date October, 2016
previous submissions 1, 2, 3, 4

 

Average results

Metric Value
IoU Classes 74.1302
iIoU Classes 52.4331
IoU Categories 89.2574
iIoU Categories 77.5879

 

Class results

Class IoU iIoU
road 98.2113 -
sidewalk 83.1775 -
building 91.5385 -
wall 44.4246 -
fence 51.1579 -
pole 63.1912 -
traffic light 70.7635 -
traffic sign 75.4867 -
vegetation 92.7329 -
terrain 70.0819 -
sky 94.5233 -
person 83.3029 66.7
rider 64.1865 46.5772
car 94.62 88.509
truck 60.8448 32.7759
bus 70.7109 41.1662
train 63.29 40.8858
motorcycle 62.9864 42.9499
bicycle 73.2437 59.9006

 

Category results

Category IoU iIoU
flat 98.4865 -
nature 92.3902 -
object 69.8057 -
sky 94.5233 -
construction 91.9183 -
human 84.2684 68.2853
vehicle 93.4091 86.8906

 

Links

Download results as .csv file

Benchmark page