Currently, the data is loaded as needed straight from the raw tif files. This is pretty inefficient. I plan to add the following functionality to speed things up.
-An np.memmap will be the primary data source for all functions within the app, thus data will be much more quickly accessible to them
-The memmap will be "dynamically filled". What I mean by that is the full scrolls are too large to populate the entire memmap at once, so I've created a custom "sub class" (not exactly) of memmap that dynamically populates it based on where the user is segmenting (and a buffer around it).
-There will be 2 options of where to populate from: local data and the download server. The download server is ideal since you can effectively segment the full dataset even if you don't have enough storage (assuming it is fast enough).
-The download server will soon be providing 500^3 blocks so that will make loading the high resolution data (for the region being segmented) easier to load quickly, as you don't have to load entire layers.
Currently, the data is loaded as needed straight from the raw tif files. This is pretty inefficient. I plan to add the following functionality to speed things up. -An np.memmap will be the primary data source for all functions within the app, thus data will be much more quickly accessible to them -The memmap will be "dynamically filled". What I mean by that is the full scrolls are too large to populate the entire memmap at once, so I've created a custom "sub class" (not exactly) of memmap that dynamically populates it based on where the user is segmenting (and a buffer around it). -There will be 2 options of where to populate from: local data and the download server. The download server is ideal since you can effectively segment the full dataset even if you don't have enough storage (assuming it is fast enough). -The download server will soon be providing 500^3 blocks so that will make loading the high resolution data (for the region being segmented) easier to load quickly, as you don't have to load entire layers.