Method Details


Details for method 'Qualcomm AI Research'

 

Method overview

name Qualcomm AI Research
challenge pixel-level semantic labeling
details Qualcomm AI Perception: More details available soon
publication Anonymous
project page / code
used Cityscapes data fine annotations
used external data ImageNet
runtime n/a
subsampling no
submission date January, 2021
previous submissions

 

Average results

Metric Value
IoU Classes 85.5667
iIoU Classes 71.1854
IoU Categories 92.9238
iIoU Categories 85.386

 

Class results

Class IoU iIoU
road 98.3439 -
sidewalk 89.1403 -
building 94.5254 -
wall 71.9927 -
fence 68.9687 -
pole 75.817 -
traffic light 82.3157 -
traffic sign 85.402 -
vegetation 93.9189 -
terrain 75.1317 -
sky 96.0626 -
person 90.0619 78.3069
rider 79.8851 63.827
car 96.9907 92.7901
truck 83.7139 60.9281
bus 95.3427 72.3685
train 88.9949 66.93
motorcycle 77.4069 62.442
bicycle 81.7529 71.8909

 

Category results

Category IoU iIoU
flat 98.3888 -
nature 93.7522 -
object 80.8846 -
sky 96.0626 -
construction 94.6006 -
human 90.1433 79.1729
vehicle 96.6347 91.5992

 

Links

Download results as .csv file

Benchmark page