Method Details
Details for method 'ML-CRNN'
Method overview
| name | ML-CRNN |
| challenge | pixel-level semantic labeling |
| details | A framework based on CNNs and RNNs is proposed, in which the RNNs are used to model spatial dependencies among image units. Besides, to enrich deep features, we use different features from multiple levels, and adopt a novel attention model to fuse them. |
| publication | Multi-level Contextual RNNs with Attention Model for Scene Labeling Heng Fan, Xue Mei, Danil Prokhorov, Haibin Ling arXiv https://arxiv.org/abs/1607.02537 |
| project page / code | |
| used Cityscapes data | fine annotations |
| used external data | ImageNet |
| runtime | n/a |
| subsampling | no |
| submission date | June, 2017 |
| previous submissions |
Average results
| Metric | Value |
|---|---|
| IoU Classes | 71.218 |
| iIoU Classes | 47.0848 |
| IoU Categories | 87.7232 |
| iIoU Categories | 72.4519 |
Class results
| Class | IoU | iIoU |
|---|---|---|
| road | 97.8648 | - |
| sidewalk | 81.0139 | - |
| building | 91.0405 | - |
| wall | 50.283 | - |
| fence | 52.4399 | - |
| pole | 56.7126 | - |
| traffic light | 65.67 | - |
| traffic sign | 71.3757 | - |
| vegetation | 92.1972 | - |
| terrain | 69.622 | - |
| sky | 94.567 | - |
| person | 80.1716 | 59.1065 |
| rider | 59.3308 | 39.0266 |
| car | 93.92 | 85.7646 |
| truck | 51.069 | 30.2938 |
| bus | 67.5987 | 40.0635 |
| train | 54.4953 | 34.1028 |
| motorcycle | 55.1275 | 34.7232 |
| bicycle | 68.643 | 53.597 |
Category results
| Category | IoU | iIoU |
|---|---|---|
| flat | 98.3405 | - |
| nature | 91.8483 | - |
| object | 64.6669 | - |
| sky | 94.567 | - |
| construction | 91.2092 | - |
| human | 80.6882 | 60.8517 |
| vehicle | 92.7421 | 84.0521 |