Method Details
Details for method 'CRFasRNN'
Method overview
| name | CRFasRNN |
| challenge | pixel-level semantic labeling |
| details | Trained on a pre-release version of the dataset |
| publication | Conditional Random Fields as Recurrent Neural Networks S. Zheng, S. Jayasumana, B. Romera-Paredes, V. Vineet, Z. Su, D. Du, C. Huang, and P. H. S. Torr ICCV 2015 http://www.cv-foundation.org/openaccess/content_iccv_2015/papers/Zheng_Conditional_Random_Fields_ICCV_2015_paper.pdf |
| project page / code | https://github.com/torrvision/crfasrnn |
| used Cityscapes data | fine annotations |
| used external data | ImageNet |
| runtime | 0.7 s CPU: Intel(R) Core(TM) i7-4960X CPU @ 3.60GHz; GPU: Nvidia Titan X. |
| subsampling | 2 |
| submission date | April, 2016 |
| previous submissions |
Average results
| Metric | Value |
|---|---|
| IoU Classes | 62.5045 |
| iIoU Classes | 34.4016 |
| IoU Categories | 82.7118 |
| iIoU Categories | 65.9932 |
Class results
| Class | IoU | iIoU |
|---|---|---|
| road | 96.283 | - |
| sidewalk | 73.8971 | - |
| building | 88.1694 | - |
| wall | 47.559 | - |
| fence | 41.2829 | - |
| pole | 35.1842 | - |
| traffic light | 49.4617 | - |
| traffic sign | 59.7282 | - |
| vegetation | 90.5596 | - |
| terrain | 66.0933 | - |
| sky | 93.4858 | - |
| person | 70.4387 | 50.6476 |
| rider | 34.6683 | 17.8203 |
| car | 90.0879 | 81.1248 |
| truck | 39.206 | 18.0111 |
| bus | 57.4719 | 24.9736 |
| train | 55.4277 | 30.2625 |
| motorcycle | 43.9473 | 22.3164 |
| bicycle | 54.6337 | 30.0565 |
Category results
| Category | IoU | iIoU |
|---|---|---|
| flat | 97.7104 | - |
| nature | 90.293 | - |
| object | 46.5066 | - |
| sky | 93.4858 | - |
| construction | 88.4782 | - |
| human | 73.5613 | 53.4027 |
| vehicle | 88.947 | 78.5837 |