Repository navigation
feature request: imshow() like matplotlib or cv2 #1702
Description
Activity
Hi @YoniChechik,
Yeah, this would be really nice. There's an open issue with the plotly.js project to add a native trace for doing this (plotly/plotly.js#3898).
One approach that I've thought of trying, but haven't yet, is overlaying a transparent heatmap on top of a layout.image. Heatmaps have an
opacityproperty that can be set to 0 while still preserving tooltips. If this was a helpful approach, we could automate it as a figure factory or plotly express utility until plotly.js adds native support.One downside, though, is that the pixels of a layout.image object are smoothed. So you don't get the crisp heatmap-style pixel boundaries when you zoom in. But this might be something that could be addressed as a new option in plotly.js for
layout.imageobjects.Are you interested in investigating this approach?
@jonmmease I'd be happy to help, but I no nothing about node js... The one input I can give is that the use of
layout.imageis wrong since the need of per-pixel separation is key! I think that changing the heatmap input to accept 3D arrays and choose each pixel color accordingly is the way to goThe one input I can give is that the use of layout.image is wrong since the need of per-pixel separation is key!
Yeah, I agree. So we'll need to wait for something to be added in the plotly.js layer.
@jonmmease check this out: plotly/plotly.js#3898 plotly/plotly.js#4289
Yep @YoniChechik we've designed
imagein plotly.js specifically to support this feature, which should come out in Plotly.py v4.3 :)
i'm familiar with go.layout.Image(), but it missing for me crucial "show closes data on hover" pixel RGB value + it can't show image from np.array.
I guess that the closest thing to start from when implementing this is heatmaps with the ability to control each pixel RGB value
it can really bring plotly to a lot more computer vision oriented programing (mainly in jupyter) and make plotly even better.