Method Details
Details for method 'SA-Gate (ResNet-101,OS=16)'
Method overview
| name | SA-Gate (ResNet-101,OS=16) |
| challenge | pixel-level semantic labeling |
| details | RGB+HHA input, input resolution = 800x800, output stride = 16, training 240 epochs, no coarse data is used. |
| publication | Bi-directional Cross-Modality Feature Propagation with Separation-and-Aggregation Gate for RGB-D Semantic Segmentation Xiaokang Chen, Kwan-Yee Lin, Jingbo Wang, Wayne Wu, Chen Qian, Hongsheng Li, and Gang Zeng European Conference on Computer Vision (ECCV), 2020 https://arxiv.org/abs/2007.09183 |
| project page / code | https://github.com/charlesCXK/RGBD_Semantic_Segmentation_PyTorch |
| used Cityscapes data | fine annotations, stereo |
| used external data | ImageNet |
| runtime | n/a |
| subsampling | no |
| submission date | August, 2020 |
| previous submissions |
Average results
| Metric | Value |
|---|---|
| IoU Classes | 82.7909 |
| iIoU Classes | 63.4636 |
| IoU Categories | 91.9453 |
| iIoU Categories | 83.0403 |
Class results
| Class | IoU | iIoU |
|---|---|---|
| road | 98.7422 | - |
| sidewalk | 87.3465 | - |
| building | 93.9224 | - |
| wall | 63.8267 | - |
| fence | 62.7393 | - |
| pole | 70.7519 | - |
| traffic light | 77.878 | - |
| traffic sign | 82.151 | - |
| vegetation | 93.9054 | - |
| terrain | 72.8117 | - |
| sky | 95.87 | - |
| person | 88.1907 | 75.2056 |
| rider | 75.169 | 55.4338 |
| car | 96.5422 | 91.5713 |
| truck | 80.4143 | 49.2184 |
| bus | 91.5955 | 59.798 |
| train | 89.0169 | 55.2118 |
| motorcycle | 73.2274 | 54.8011 |
| bicycle | 78.9256 | 66.4692 |
Category results
| Category | IoU | iIoU |
|---|---|---|
| flat | 98.7721 | - |
| nature | 93.5616 | - |
| object | 76.7597 | - |
| sky | 95.87 | - |
| construction | 94.1338 | - |
| human | 88.4278 | 76.0918 |
| vehicle | 96.0919 | 89.9888 |