Method Details
Details for method 'Learnable Tree Filter V2'
Method overview
| name | Learnable Tree Filter V2 |
| challenge | pixel-level semantic labeling |
| details | Based on ResNet-101 backbone and FPN architecture. |
| publication | Rethinking Learnable Tree Filter for Generic Feature Transform Lin Song, Yanwei Li, Zhengkai Jiang, Zeming Li, Xiangyu Zhang, Hongbin Sun, Jian Sun, Nanning Zheng NeurIPS 2020 https://papers.nips.cc/paper/2020/file/2952351097998ac1240cb2ab7333a3d2-Paper.pdf |
| project page / code | https://github.com/StevenGrove/LearnableTreeFilterV2 |
| used Cityscapes data | fine annotations |
| used external data | ImageNet |
| runtime | n/a |
| subsampling | no |
| submission date | November, 2019 |
| previous submissions |
Average results
| Metric | Value |
|---|---|
| IoU Classes | 82.1409 |
| iIoU Classes | 63.953 |
| IoU Categories | 92.0937 |
| iIoU Categories | 83.5546 |
Class results
| Class | IoU | iIoU |
|---|---|---|
| road | 98.657 | - |
| sidewalk | 86.5303 | - |
| building | 93.6476 | - |
| wall | 57.4501 | - |
| fence | 61.1811 | - |
| pole | 71.863 | - |
| traffic light | 79.5768 | - |
| traffic sign | 82.6756 | - |
| vegetation | 93.9868 | - |
| terrain | 72.4591 | - |
| sky | 95.8644 | - |
| person | 88.482 | 76.3294 |
| rider | 75.3175 | 57.3712 |
| car | 96.5647 | 91.5103 |
| truck | 76.1941 | 49.1907 |
| bus | 88.3554 | 58.3571 |
| train | 87.8647 | 56.6304 |
| motorcycle | 74.4506 | 54.3555 |
| bicycle | 79.5556 | 67.8792 |
Category results
| Category | IoU | iIoU |
|---|---|---|
| flat | 98.7416 | - |
| nature | 93.6326 | - |
| object | 77.5418 | - |
| sky | 95.8644 | - |
| construction | 94.1133 | - |
| human | 88.737 | 77.3254 |
| vehicle | 96.0251 | 89.7838 |