Method Details
Details for method 'PL-Seg'
Method overview
| name | PL-Seg |
| challenge | pixel-level semantic labeling |
| details | Following "partial order pruning", we conduct architecture searching experiments on Snapdragon 845 platform, and obtained PL1A/PL1A-Seg. 1、Snapdragon 845 2、NCNN Library 3、latency evaluated at 640x384 |
| publication | Partial Order Pruning: for Best Speed/Accuracy Trade-of in Neural Architecture Search Xin Li, Yiming Zhou, Zheng Pan, Jiashi Feng CVPR 2019 https://arxiv.org/abs/1903.03777 |
| project page / code | https://github.com/lixincn2015/Partial-Order-Pruning |
| used Cityscapes data | fine annotations |
| used external data | ImageNet |
| runtime | 0.0192 s Snapdragon 845 |
| subsampling | no |
| submission date | February, 2019 |
| previous submissions |
Average results
| Metric | Value |
|---|---|
| IoU Classes | 69.0972 |
| iIoU Classes | 41.2344 |
| IoU Categories | 86.3961 |
| iIoU Categories | 67.6951 |
Class results
| Class | IoU | iIoU |
|---|---|---|
| road | 97.8703 | - |
| sidewalk | 80.8257 | - |
| building | 90.1723 | - |
| wall | 41.9726 | - |
| fence | 44.3962 | - |
| pole | 52.2669 | - |
| traffic light | 59.8478 | - |
| traffic sign | 66.3198 | - |
| vegetation | 91.7776 | - |
| terrain | 68.7447 | - |
| sky | 94.6633 | - |
| person | 77.8161 | 51.4537 |
| rider | 56.8869 | 32.4667 |
| car | 93.2024 | 85.0521 |
| truck | 54.3834 | 22.0632 |
| bus | 67.6704 | 35.9413 |
| train | 60.6119 | 30.1587 |
| motorcycle | 48.4229 | 24.7127 |
| bicycle | 64.995 | 48.0265 |
Category results
| Category | IoU | iIoU |
|---|---|---|
| flat | 98.2449 | - |
| nature | 91.427 | - |
| object | 59.8367 | - |
| sky | 94.6633 | - |
| construction | 90.3932 | - |
| human | 78.1854 | 52.8687 |
| vehicle | 92.022 | 82.5216 |