Trying to run with our data

[albertmena]

Hello and thanks for developing that fantastic software. We at the Spanish National Centre of Biotechnolog are trying to test it with our data and we have some issues.

• Does the network support different meshes? (grid mesh 200 300 400, R0.6/1, R2/2...)
• What do you mean by low-mag and med-mag? (x80 and x1000?)
• What is the pixelSize of the network input images?

Thanks

[ptkim1]

Thanks for your interest!

• The model supports a wide range of pixelsizes, as it was trained on data from different microscopes using different pixel sizes. I would assume that it works on all reasonable pixel sizes, until we see evidence otherwise.
• Our algorithms should be agnostic to grid mesh. If you try them out and observe different results please let us know!
• In general, low-mag means a magnification at the grid level, where many squares in the grid are visible simultaneously. Similarly, medium mag means a magnification within squares where many holes are visible simultaneously. We don't have a fixed magnification for either of these levels. The only restriction is that if the holes in the medium mag are too large (in terms of number of pixels), the model might struggle to detect the hole centers properly - this can be fixed by simply downsampling the medium mag image.

Thanks