Nir Hutnik

### Inputs: ignore_na: exactly as in drift checks ### Logic: * if ignore_na is True - drop them from each column * if ignore_na is False - use astype(str) to...

### Tasks: * Search for sources - 2 SPs * Create list of possible candidates and expand by algo brainstorm - 1 SP * Prioritize list and show top candidates...

Closes #1887

So, as @matanper showed, the FeatureLabelCorrelation check (that uses the ppscore method) does not take into account Nones. This is because it is based on training simple models to find...

### Background: In LabelPropertyCorrelationChange and PropertyLabelCorrelation, we need to associate image properties with a label. Therefor, for object detection, we crop out the bounding box and send it to the...

Just change the wording to clarify that this custom check is instance segmentation and not semantic segmentation And change the class names to reflect this

Out of the above suggestions, implement: * Hover-on: add the contribution of that class to the drift Do not implement: * Hover-on: add the average score on this prediction class...

I undertstand this comes from this agenda. But - 1. Performance on train is not representative of test and can be misleading 2. Even so, performance on each category individually...

1. If model is overfitted, the estimation will be very wrong. If you had validation set I would have agreed, but training alone is prone to huge errors. 2. Better...

### Description: Proposed solution - use regular [plotly.graph_objects.Heatmap](https://plotly.com/python-api-reference/generated/plotly.graph_objects.Heatmap.html) object where the parameter connectgaps can be used and turned to False. To my understanding, it makes the graph [look like this](https://chart-studio.plotly.com/~plotly.js/169.embed)....