Table 1 Results of our underwater segmentation models, averaged over the 300 test images
From: Underwater image segmentation in the wild using deep learning
Architect. | Pre-processed | PASCAL Pre-trained | Training data | mIoU |
|---|---|---|---|---|
SegNet | UDCP | Yes | - | 0.415 |
SegNet | UGAN | Yes | - | 0.421 |
SegNet | - | No | NAUTEC UWI Real | 0.825 |
SegNet | - | No | NAUTEC UWI Mixed | 0.805 |
SegNet | - | No | NAUTEC UWI Sim200 | 0.522 |
SegNet | - | No | NAUTEC UWI Sim1000 | 0.565 |
SegNet | - | Yes | - | 0.444 |
SegNet | - | Yes | NAUTEC UWI Real | 0.796 |
SegNet | - | Yes | NAUTEC UWI Mixed | 0.795 |
SegNet | - | Yes | NAUTEC UWI Sim200 | 0.558 |
SegNet | - | Yes | NAUTEC UWI Sim1000 | 0.556 |
DeepLab | UDCP | Yes | - | 0.434 |
DeepLab | UGAN | Yes | - | 0.485 |
DeepLab | - | No | NAUTEC UWI Real | 0.689 |
DeepLab | - | No | NAUTEC UWI Mixed | 0.617 |
DeepLab | - | No | NAUTEC UWI Sim200 | 0.445 |
DeepLab | - | No | NAUTEC UWI Sim1000 | 0.481 |
DeepLab | - | Yes | - | 0.444 |
DeepLab | - | Yes | NAUTEC UWI Real | 0.919 |
DeepLab | - | Yes | NAUTEC UWI Mixed | 0.909 |
DeepLab | - | Yes | NAUTEC UWI Sim200 | 0.762 |
DeepLab | - | Yes | NAUTEC UWI Sim1000 | 0.751 |