dianna

dianna copied to clipboard

Unclear error message when mistyping a method name

6

  It should have given the valid options as well (RISE, LIME, kernelSHAP).

cwmeijer

Currently we are masking images with zeros which is not always sensible. The behavior will be very different for a bright image as opposed to a dark image I would...

cwmeijer

Rise outputs a numpy array with dimensions n_labels, height, width, while Lime outputs a list of numpy arrays with dimensions height, width. It should both output the same data structure.

cwmeijer

KernelSHAP doesn't accept model functions

1

It only seems to work with model objects which is inconsistent with how the other methods work. We always want to support model functions and model objects. We should think...

cwmeijer

Enable DIANNA to work with (image) models which take more than 1 input.

2

DIANNA should be able to run on models with more than 1 input. Inspired by the Lorentz workshop CNN-post-processing use case. For example data and quick fix, see the link...

elboyran

Follows from #406

elboyran

Enable DIANNA to take images with multiple channels as input

2

DIANNA should be able to process images as inputs and models trained with more than 3 (e.g. RGB) channels. Inspired by the Lorentz workshop CNN-post-processing use case. For example data...

elboyran

follows from #406 Related to the prepared model for the Coffee dataset (#385)

elboyran

follows from #406

elboyran

Make design for multiple text items explained side-by-side

1

cpranav93