Hi formfrei,
I have two ways to do it, have a look. The first one discusses a little bit more the problems. With that said, later I show a very simple set up to get the data.
Baking the information for further refinement before use.
Some notes about this:
Use floating point formats to do so (Open EXR, Tiff 32bit/channel, etc. Radiance [HDR] is not a quality format and has only 4*8bit, avoid it!)
Set the mid gray point to 0.18, as this is the only possible stable value. There is no black or white point in HDRI; 100% is not white, it is where integer is clipped and lost all color information)
Set a max brightness when things might clip, e.g., 4.0 (400% in integer lingo), and normalize these values to 1.0 max, as the weight map can’t handle values over 1.0 to my knowledge (or 100%)
The Weight maps have no need for UV(W) data, as they are based on the vertices’ ID or “point-number”. So they store only that specific value.
The data taken from an image will of course be translated to sit on the vertices, if via projection or UV(W).
Some notes here:
The value taken from that image is mostly based, for a specific point from a small area of the image.
An image could be blurred or reduced in resolution to provide a more balanced read-out.
The blurred, averaged or down-sampled value will flatten the overall dynamic.
The simplest way would be to use the Shader Effector, and take the information given to the clones,
Scene File
https://www.amazon.com/clouddrive/share/rnHIGmxTL61ocG1QcLC5XeiaVB3OvDMj4fr8Qw4S5Fy?ref_=cd_ph_share_link_copy
The live version
With that all said and done, the base concerns about the representation of light stored in vertex maps can start with Thinking Particles.
Placing particles on each vertices and using the PLight.
Scene file:
https://www.amazon.com/clouddrive/share/gUeY3b7b0QMRbW0tVuyG0VHbcfpzhBibTteEjnOuKZk?ref_=cd_ph_share_link_copy
The particles could store the values in their channels, and being used then later for what ever reasons. Weight stores real values, MoGraph selection can store “boole” values, as in light on or not.
A simpler way would be to ignore particles and take only the PLight node, which needs only the position and light to evaluate the set up
Scene file:
https://www.amazon.com/clouddrive/share/IOlsC8Rlw9P3H1TncXeAJ4NCFfZkJYUY5FZST5U8fXm?ref_=cd_ph_share_link_copy
You might check out if Color or Intensity needs to be multiplied and normalized, as discussed above. Here is an option just color * intensity
Scene file:
https://www.amazon.com/clouddrive/share/OW2MwQDLxtkpcxzDIOxePFzjIUpWlz5ydMObyX89ZtH?ref_=cd_ph_share_link_copy
I placed the copy of that object into the scene to show the effect.
Screen-shot:
https://www.amazon.com/clouddrive/share/o77igWMXwcIjygH3vdzxk53JgBmYSNzpGr18NdulII?ref_=cd_ph_share_link_copy
All the best
P.S.: Since the position of the vertices doesn’t care about shadows, you might use the Normal Direction of the points to improve the set up. /edit