Pipeline for labeling training data and training networks

[esteinig]

Pipeline of collecting images, labeling cell types / parasites (Leaflet) and training networks, how would we implement this link to other modules in Autoscope?

[Wytamma]

This might be useful for labelling data.

[esteinig]

Leaflet would be great for targeting parasites and counting / binary classification. If we want to extend to cell types or so later ,we might want to implement bounding boxes?

[Wytamma]

Yeah I think that’s a good idea. If we convert the slide scans into map tiles we can use them with leaflet (see gdal2tiles.py). We can attach a simple database to the example above and use the leaflet polygon feature to add bounding boxes. Some simple export code could get it from the DB to a ML image data format.

[esteinig]

Yes, was thinking about the polygon feature, that should be fairly simple. Will check this out when I am out of the lab.