Method Details
Details for method 'RefineNet'
Method overview
| name | RefineNet |
| challenge | pixel-level semantic labeling |
| details | Please refer to our technical report for details: "RefineNet: Multi-Path Refinement Networks for High-Resolution Semantic Segmentation" (https://arxiv.org/abs/1611.06612). Our source code is available at: https://github.com/guosheng/refinenet 2975 images (training set with fine labels) are used for training. |
| publication | RefineNet: Multi-Path Refinement Networks for High-Resolution Semantic Segmentation Guosheng Lin; Anton Milan; Chunhua Shen; Ian Reid; https://arxiv.org/abs/1611.06612 |
| project page / code | https://github.com/guosheng/refinenet |
| used Cityscapes data | fine annotations |
| used external data | ImageNet |
| runtime | n/a |
| subsampling | no |
| submission date | November, 2016 |
| previous submissions |
Average results
| Metric | Value |
|---|---|
| IoU Classes | 73.6015 |
| iIoU Classes | 47.1978 |
| IoU Categories | 87.9163 |
| iIoU Categories | 70.6453 |
Class results
| Class | IoU | iIoU |
|---|---|---|
| road | 98.2027 | - |
| sidewalk | 83.3112 | - |
| building | 91.2843 | - |
| wall | 47.7873 | - |
| fence | 50.4031 | - |
| pole | 56.1162 | - |
| traffic light | 66.922 | - |
| traffic sign | 71.3 | - |
| vegetation | 92.282 | - |
| terrain | 70.3261 | - |
| sky | 94.7591 | - |
| person | 80.879 | 55.566 |
| rider | 63.2806 | 35.7979 |
| car | 94.5104 | 86.9125 |
| truck | 64.5625 | 30.0536 |
| bus | 76.0794 | 42.6215 |
| train | 64.2705 | 42.3803 |
| motorcycle | 62.2009 | 34.2915 |
| bicycle | 69.9521 | 49.9593 |
Category results
| Category | IoU | iIoU |
|---|---|---|
| flat | 98.4015 | - |
| nature | 91.9335 | - |
| object | 63.7734 | - |
| sky | 94.7591 | - |
| construction | 91.7009 | - |
| human | 81.2514 | 56.8031 |
| vehicle | 93.5942 | 84.4875 |