Method Details
Details for method 'ESANet RGB-D (small input)'
Method overview
| name | ESANet RGB-D (small input) |
| challenge | pixel-level semantic labeling |
| details | Efficient RGB-D Semantic Segmentation for Indoor Scene Analysis. ESANet-R34-NBt1D using RGB-D data with half the input resolution. |
| publication | Efficient RGB-D Semantic Segmentation for Indoor Scene Analysis Daniel Seichter, Mona Köhler, Benjamin Lewandowski, Tim Wengefeld and Horst-Michael Gross |
| project page / code | https://github.com/TUI-NICR/ESANet |
| used Cityscapes data | fine annotations, stereo |
| used external data | |
| runtime | 0.0427 s NVIDIA Jetson AGX Xavier (Jetpack 4.4, TensorRT 7.1, Float16) |
| subsampling | 2 |
| submission date | October, 2020 |
| previous submissions |
Average results
| Metric | Value |
|---|---|
| IoU Classes | 75.646 |
| iIoU Classes | 44.908 |
| IoU Categories | 88.9559 |
| iIoU Categories | 70.6886 |
Class results
| Class | IoU | iIoU |
|---|---|---|
| road | 98.4995 | - |
| sidewalk | 85.9197 | - |
| building | 92.3479 | - |
| wall | 54.0931 | - |
| fence | 55.4349 | - |
| pole | 61.6331 | - |
| traffic light | 67.9727 | - |
| traffic sign | 72.8438 | - |
| vegetation | 92.3085 | - |
| terrain | 71.3468 | - |
| sky | 95.1956 | - |
| person | 82.4766 | 56.059 |
| rider | 65.8011 | 35.4149 |
| car | 94.8204 | 87.424 |
| truck | 64.0218 | 22.96 |
| bus | 79.9297 | 38.1479 |
| train | 72.3218 | 39.5896 |
| motorcycle | 60.4894 | 32.0618 |
| bicycle | 69.8169 | 47.607 |
Category results
| Category | IoU | iIoU |
|---|---|---|
| flat | 98.4663 | - |
| nature | 92.0047 | - |
| object | 67.9954 | - |
| sky | 95.1956 | - |
| construction | 92.3724 | - |
| human | 82.43 | 56.7565 |
| vehicle | 94.2269 | 84.6208 |