dianna
dianna copied to clipboard
dianna-ai
Deep Insight And Neural Network Analysis
We support images with 1 or 3 channels, but no grayscale images without channels axis. This would be nice to support. See also the discussion in issue #222.
While we were working on #188, @loostrum noticed a bug in the first commit 886dbd776577bdad6a6e5abd01510acfc115cb92, which has `if method == "SHAP"`, which should have been `KernelSHAP`. We fixed it in...
To avoid having to edit the contents of docs/tutorials when adding/changing a tutorial notebook, we could automatically generate the files in docs/tutorials as part of the doc building CI.
Right now, in #74, docstrings were introduced that fulfill all the technical requirements from linting. However, these requirements in some cases do not yield satisfactory results. For instance, methods.LIME.__init__ lists...
Output of rise can be added repeatedly so you don't have to know the final n_masks in advance. We can keep adding masks until the output doesn't change much anymore....
Create text visualization: ordered table with weights
At the moment there are a number of hardcoded things for the imagenet model in the dashboard. One of them is related to the fact that the Resnet model requires...
Following #201, we should also replace use of the term heatmap in code, where relevant. After #210, some 50 uses of "heatmap" remain, which can be found with the following...
- [ ] Center-wide and/or at least the ML SIG? - [ ] The people who contributed to the user stories - [ ] List of our past and future...
