Jacob Rosenthal
Jacob Rosenthal
Thanks for making this great tool!!! I just noticed that on the website, the dates from the table and the graph are off by one. See the screenshot attached, where...
For some operations, e.g. tissue detection HE, we may want to run at lower resolution to speed up computation relative to running on each tile at full-resolution. What is the...
Right now `QuantifyMIF` is used for taking a cell segmentation mask and computing cell-level statistics into a counts matrix. However, there is no reason why this should be limited to...
We should have a clean and tested integration with [deepcell](https://github.com/vanvalenlab/deepcell-tf). For example, we could add a Transform which wraps a deepcell [`application`](https://deepcell.readthedocs.io/en/master/API/deepcell.applications.html). This would basically generalize our current implementation of...
Add U-net, plus tests. This is the model that they used in the PESO paper
This is a relatively small (
We want to be able to share pre-trained models. The trained model weights can be saved to disk, e.g. in `.pth` files for pytorch. However, these files can be quite...
We can define a `SlideData._repr_html_()` method (or maybe `SlideData._repr_jpg_()`) which would let us do pretty outputs in JupyterNotebook. For example we could make this method display a thumbnail of the...
Add some tests to make sure that the ML models are working correctly. For example, this may involve overfitting on a toy dataset and verifying that performance is above some...
A new version of openslide was recently released with some improvements, notably support for DICOM. Maybe pathml support for DICOM could be moved to use openslide instead of the backend...