Image normalization to account for lighting conditions, by using a color reference zone
Is your feature request related to a problem? Please describe. No
Describe the solution you'd like Would love the ability to set a color reference zone. For example, the side of the house, or the white trim around a window. Essentially, a region that should be used to modify the values to account for lighting conditions. When the reference zone is created, the colors in the zone would be averaged, stored permanently and used as a reference to determine lighting conditions for the image in future detections. Suggest a minimum of 25 pixels for this area. They could even just be individual pixels from parts of the image that won't be used for object detection.
For example, if your reference zone averages to rgb(117, 128, 131) when created, it would be checked again before object detection is processed. If it now averages to rgb(129, 127, 122), each channel in the object detection zone would be multiplied by some constant to normalize the colors - in this case the constants would be rgb(0.907,1.008,1.074).
Another thing to note is that colors should be normalized when saving the color variant for an object.
Describe alternatives you've considered Maybe implementing this myself? But I'd need at least the ability to template the colors of variants and am not sure if that's possible without having to constantly reload the configuration.
Additional context I literally haven't even implemented this integration yet 😃, so take this with a grain of salt. Just something I considered while reading the docs.
That would definitely be a good idea, although seems pretty complicated. I'll have to think about how I might be able to do that.