Holoviews: How to wrap multiple RGB images into a Dataset and display them correctly?

Question

I have a dataset of RGB images as a numpy array of shape (n, height, width, 3). I want to wrap the data into a hv.Dataset and visualize it (with bokeh) such that:

• A slider can be used to select one of the n images.
• Each image is displayed in its correct aspect ratio and orientaton.

I'm primarily interested in a solution without xarray.

Answer 1

Something like this should work, though I think there's probably a cleaner approach declaring an hv.Dataset wrapping the input array:

import numpy as np, holoviews as hv
hv.extension('bokeh')

x,y = np.mgrid[-50:51, -100:101] * 0.1

def ms(k):
    r = 0.5*np.sin(np.pi  +k*x**2+y**2)+0.5
    g = 0.5*np.sin(x**2+k*y**2)+0.5
    b = 0.5*np.sin(np.pi/k+x**2+y**2)+0.5
    return np.dstack([r,g,b])

a = np.stack([ms(i) for i in [1,2,3,4]])

hv.HoloMap({i:hv.RGB(a[i-1,:,:,:], bounds=(-1,-0.5,1,0.5)).opts(data_aspect=1) for i in [1,2,3,4]})