Method Details
Details for method 'Pixel-level Encoding for Instance Segmentation'
Method overview
| name | Pixel-level Encoding for Instance Segmentation |
| challenge | pixel-level semantic labeling |
| details | We predict three encoding channels from a single image using an FCN: semantic labels, depth classes, and an instance-aware representation based on directions towards instance centers. Using low-level computer vision techniques, we obtain pixel-level and instance-level semantic labeling paired with a depth estimate of the instances. |
| publication | Pixel-level Encoding and Depth Layering for Instance-level Semantic Labeling J. Uhrig, M. Cordts, U. Franke, and T. Brox GCPR 2016 http://arxiv.org/abs/1604.05096 |
| project page / code | |
| used Cityscapes data | fine annotations, stereo |
| used external data | ImageNet |
| runtime | n/a |
| subsampling | no |
| submission date | April, 2016 |
| previous submissions |
Average results
| Metric | Value |
|---|---|
| IoU Classes | 64.3053 |
| iIoU Classes | 41.5847 |
| IoU Categories | 85.8862 |
| iIoU Categories | 73.8804 |
Class results
| Class | IoU | iIoU |
|---|---|---|
| road | 97.3514 | - |
| sidewalk | 77.7158 | - |
| building | 88.764 | - |
| wall | 27.7383 | - |
| fence | 40.1327 | - |
| pole | 51.4714 | - |
| traffic light | 60.0827 | - |
| traffic sign | 64.6672 | - |
| vegetation | 91.1441 | - |
| terrain | 67.6334 | - |
| sky | 93.5146 | - |
| person | 77.737 | 60.5757 |
| rider | 54.1634 | 33.4172 |
| car | 92.4148 | 86.7287 |
| truck | 33.681 | 19.518 |
| bus | 41.9876 | 25.6202 |
| train | 42.5086 | 25.7525 |
| motorcycle | 52.5489 | 30.5481 |
| bicycle | 66.5438 | 50.5171 |
Category results
| Category | IoU | iIoU |
|---|---|---|
| flat | 98.1834 | - |
| nature | 90.7538 | - |
| object | 59.3101 | - |
| sky | 93.5146 | - |
| construction | 89.1592 | - |
| human | 79.1525 | 62.5841 |
| vehicle | 91.1298 | 85.1767 |