Method Details
Details for method 'YOLO V5s with Segmentation Head'
Method overview
| name | YOLO V5s with Segmentation Head |
| challenge | pixel-level semantic labeling |
| details | Multitask model. fine tune from COCO detection pretrained model, train semantic segmentation and object detection(transfer from instance label) at the same time |
| publication | Anonymous |
| project page / code | https://github.com/TomMao23/multiyolov5 |
| used Cityscapes data | fine annotations |
| used external data | COCO detection pretrained model, train sematic segmentation and detection(transfer from instance label) at the same time |
| runtime | 0.007 s 2080Ti |
| subsampling | 2 |
| submission date | May, 2021 |
| previous submissions |
Average results
| Metric | Value |
|---|---|
| IoU Classes | 71.3012 |
| iIoU Classes | 46.2572 |
| IoU Categories | 85.6674 |
| iIoU Categories | 70.3735 |
Class results
| Class | IoU | iIoU |
|---|---|---|
| road | 97.9575 | - |
| sidewalk | 81.2516 | - |
| building | 90.2368 | - |
| wall | 41.9417 | - |
| fence | 44.8716 | - |
| pole | 44.7686 | - |
| traffic light | 62.6257 | - |
| traffic sign | 67.7752 | - |
| vegetation | 90.7369 | - |
| terrain | 67.403 | - |
| sky | 93.4803 | - |
| person | 78.4086 | 55.977 |
| rider | 61.6029 | 36.3805 |
| car | 94.1188 | 86.137 |
| truck | 65.3371 | 28.5333 |
| bus | 77.1281 | 38.8018 |
| train | 70.2055 | 40.8572 |
| motorcycle | 58.5339 | 32.8679 |
| bicycle | 66.3397 | 50.503 |
Category results
| Category | IoU | iIoU |
|---|---|---|
| flat | 98.2061 | - |
| nature | 90.4172 | - |
| object | 55.6085 | - |
| sky | 93.4803 | - |
| construction | 90.2644 | - |
| human | 78.5252 | 57.0658 |
| vehicle | 93.17 | 83.6813 |