journal menu
![[Figure 1]](jo5091fig1mag.jpg)
|
|
Figure 1
Schematic of the pipeline for training the pre-trained Mask R-CNN model (He et al., 2017  ) on our target G. Mask R-CNN adds a branch for object mask prediction, in parallel to bounding box recognition. The model generates bounding boxes, provides the objectness score for detection and accurately segments the intracellular crystals, requiring only a few annotated images as the training dataset. |
![[Figure 1]](jo5091fig1.jpg)
 | JOURNAL OF APPLIED CRYSTALLOGRAPHY |
ISSN: 1600-5767
access
