Abstract | ||
---|---|---|
•We introduce the idea of local rotation invariance in 3D convolutional networks.•We propose three implementations including the use of spherical harmonics and 3D steerable filters.•We demonstrate the benefit of these approaches on a synthetic 3D texture dataset and a dataset of 3D CT scans of pulmonary nodules.•An important aspect of the reduction of trainable parameters is obtained from the weight sharing and the parametric representation of the 3D filters. |
Year | DOI | Venue |
---|---|---|
2020 | 10.1016/j.media.2020.101756 | Medical Image Analysis |
Keywords | DocType | Volume |
Local rotation invariance,Convolutional neural network,Steerable filters,3D Texture | Journal | 65 |
ISSN | Citations | PageRank |
1361-8415 | 0 | 0.34 |
References | Authors | |
23 | 5 |
Name | Order | Citations | PageRank |
---|---|---|---|
Vincent Andrearczyk | 1 | 3 | 1.39 |
Julien Fageot | 2 | 13 | 6.42 |
Oreiller Valentin | 3 | 0 | 0.34 |
Montet Xavier | 4 | 0 | 0.34 |
Adrien Depeursinge | 5 | 418 | 38.83 |