Deep learning segmentation model; can be loaded in the feature configuration (F3) -> Deep Learning Segmentation tab, and used for cell segmentation on both, fluorescence and brightfield images:
This model is based of the third place winner model of the Kaggle 2018 Data Science Bowl cell segmentation challenge.
Please note that the out-of-the-box deep learning segmentation only uses CPU and is pretty slow. If you want to use this in a production environment, I highly recommend to use a computer with (Nvidia) GPU, checkout the Orbit source-code, activate tensorflow-gpu in the build.gradle and re-build Orbit or simply run the InstSegMaskRCNN class on your own. Please let me know if you need any help for this.
Deep learning models to be used with the deep learning groovy script (to be executed in the Tools -> Script Editor) to detect arbitrary heterogeneous objects:
- Glomeruli detection for H/DAB staining in kidney (rat or mouse, works with other Hematoxylin/x staining as well)
If you want to train our own deep learning models, you can use the python scripts for it.