Method Details
Details for method 'K-net'
Method overview
name | K-net |
challenge | pixel-level semantic labeling |
details | |
publication | XinLiang Zhong |
project page / code | |
used Cityscapes data | fine annotations |
used external data | |
runtime | n/a |
subsampling | no |
submission date | October, 2017 |
previous submissions |
Average results
Metric | Value |
---|---|
IoU Classes | 76.0295 |
iIoU Classes | 52.8334 |
IoU Categories | 88.792 |
iIoU Categories | 75.4128 |
Class results
Class | IoU | iIoU |
---|---|---|
road | 98.3039 | - |
sidewalk | 84.301 | - |
building | 92.0653 | - |
wall | 52.2809 | - |
fence | 56.4736 | - |
pole | 58.9971 | - |
traffic light | 69.8376 | - |
traffic sign | 73.1501 | - |
vegetation | 92.654 | - |
terrain | 70.3813 | - |
sky | 94.6452 | - |
person | 82.9531 | 62.7581 |
rider | 66.2565 | 46.6973 |
car | 95.2292 | 88.3082 |
truck | 68.5006 | 36.9815 |
bus | 79.924 | 51.716 |
train | 73.9987 | 42.5901 |
motorcycle | 62.5426 | 38.5722 |
bicycle | 72.0659 | 55.0439 |
Category results
Category | IoU | iIoU |
---|---|---|
flat | 98.4508 | - |
nature | 92.3149 | - |
object | 66.3888 | - |
sky | 94.6452 | - |
construction | 92.3713 | - |
human | 82.9618 | 64.5661 |
vehicle | 94.4111 | 86.2595 |