This is a segmentation network to classify each pixel into 20 classes:
| Metric | Value |
|---|---|
| Image size | 2048x1024 |
| GFlops | 58.572 |
| MParams | 6.686 |
| Source framework | PyTorch* |
The quality metrics calculated on 2000 images:
| Label | IOU |
|---|---|
| mean | 0.6907 |
| Road | 0.910379 |
| Sidewalk | 0.630676 |
| Building | 0.860139 |
| Wall | 0.424166 |
| Fence | 0.592632 |
| Pole | 0.559078 |
| Traffic Light | 0.654779 |
| Traffic Sign | 0.648217 |
| Vegetation | 0.882593 |
| Terrain | 0.620521 |
| Sky | 0.976889 |
| Person | 0.711653 |
| Rider | 0.612787 |
| Car | 0.877892 |
| Truck | 0.674829 |
| Bus | 0.743752 |
| Train | 0.358641 |
| Motorcycle | 0.600701 |
| Bicycle | 0.622246 |
| Ego-Vehicle | 0.852932 |
IOU=TP/(TP+FN+FP), where:TP - number of true positive pixels for given classFN - number of false negative pixels for given classFP - number of false positive pixels for given classThe blob with BGR image in format: [B, C=3, H=1024, W=2048], where:
[*] Other names and brands may be claimed as the property of others.