Name
Abstract | ||
|---|---|---|
•Impose inequality constraints during training of neural networks for semantic segmentation.•Constraints are based on anatomical priors (shape, size, ...).•Enable the use of weak annotations such as dots, scribbles.•Add a penalty directly into the loss function, which is very simple and efficient.•Demonstrate the effectiveness on three applications (left-ventricle, vertebral-body, prostate segmentation.•Get close to full supervision performances with 0.1% of labeled pixels Code is publicly available. |
Year | DOI | Venue |
|---|---|---|
2019 | 10.1016/j.media.2019.02.009 | Medical Image Analysis |
Keywords | DocType | Volume |
Deep learning,Semantic segmentation,Weakly-supervised learning,CNN constraints | Journal | 54 |
ISSN | Citations | PageRank |
1361-8415 | 11 | 0.67 |
References | Authors | |
0 | 6 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Hoel Kervadec | 1 | 25 | 2.94 |
| Dolz Jose | 2 | 91 | 16.76 |
| Meng Tang | 3 | 123 | 6.88 |
| E Granger | 4 | 62 | 7.10 |
| Yuri Boykov | 5 | 7601 | 497.20 |
| Ismail Ben Ayed | 6 | 678 | 52.28 |