Method Details
Details for method 'F2MF-short'
Method overview
| name | F2MF-short |
| challenge | pixel-level semantic labeling |
| details | Our method forecasts semantic segmentation 3 timesteps into the future. |
| publication | Warp to the Future: Joint Forecasting of Features and Feature Motion Josip Saric, Marin Orsic, Tonci Antunovic, Sacha Vrazic, Sinisa Segvic The IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020 http://openaccess.thecvf.com/content_CVPR_2020/html/Saric_Warp_to_the_Future_Joint_Forecasting_of_Features_and_Feature_CVPR_2020_paper.html |
| project page / code | https://jsaric.github.io/f2mf/ |
| used Cityscapes data | fine annotations, video |
| used external data | ImageNet |
| runtime | n/a |
| subsampling | no |
| submission date | November, 2019 |
| previous submissions |
Average results
| Metric | Value |
|---|---|
| IoU Classes | 70.2134 |
| iIoU Classes | 43.6408 |
| IoU Categories | 82.4503 |
| iIoU Categories | 61.5138 |
Class results
| Class | IoU | iIoU |
|---|---|---|
| road | 97.2557 | - |
| sidewalk | 78.7326 | - |
| building | 88.9778 | - |
| wall | 54.0235 | - |
| fence | 52.0547 | - |
| pole | 46.0101 | - |
| traffic light | 57.9649 | - |
| traffic sign | 61.8967 | - |
| vegetation | 89.5761 | - |
| terrain | 66.1592 | - |
| sky | 91.634 | - |
| person | 65.933 | 46.3204 |
| rider | 54.2409 | 31.252 |
| car | 90.0327 | 77.6609 |
| truck | 66.5071 | 30.4042 |
| bus | 80.7245 | 43.0473 |
| train | 76.9637 | 46.9537 |
| motorcycle | 54.0241 | 30.4221 |
| bicycle | 61.3435 | 43.0661 |
Category results
| Category | IoU | iIoU |
|---|---|---|
| flat | 97.2902 | - |
| nature | 89.2284 | - |
| object | 54.4639 | - |
| sky | 91.634 | - |
| construction | 89.1635 | - |
| human | 66.2003 | 47.0782 |
| vehicle | 89.1717 | 75.9494 |