Method Details


Details for method 'BiseAsspDcnDts-v1'

 

Method overview

name BiseAsspDcnDts-v1
challenge pixel-level semantic labeling
details use ASSP、DCN、Depth to space model for semantic segmentation
publication Anonymous
project page / code
used Cityscapes data fine annotations
used external data
runtime 0.0128 s
NVIDIA Tesla V100 32GB
subsampling no
submission date September, 2020
previous submissions

 

Average results

Metric Value
IoU Classes 74.8585
iIoU Classes 53.491
IoU Categories 90.1755
iIoU Categories 78.6611

 

Class results

Class IoU iIoU
road 98.4404 -
sidewalk 85.0725 -
building 92.3015 -
wall 48.4282 -
fence 55.3905 -
pole 65.0352 -
traffic light 73.975 -
traffic sign 76.1325 -
vegetation 93.2731 -
terrain 71.232 -
sky 94.9386 -
person 85.3081 68.4646
rider 67.295 45.6909
car 95.4474 89.3784
truck 57.8666 30.7824
bus 67.5854 49.1795
train 58.8216 41.4946
motorcycle 61.3478 40.6074
bicycle 74.4195 62.3303

 

Category results

Category IoU iIoU
flat 98.6433 -
nature 92.9725 -
object 71.2619 -
sky 94.9386 -
construction 92.7825 -
human 85.792 69.8502
vehicle 94.8376 87.4721

 

Links

Download results as .csv file

Benchmark page