Going from high bit depth linear data to 8 bits per pixel true color imagery is by no means straightforward, but here is a tutorial that might help you get started: https://medium.com/@robsimmon/making-sense-of-satellite-data-an-open-source-workflow-stitching-data-with-qgis-d004c2184ebb

Hi all, Rebecca’s supervisor here to elaborate on her (well-posed!) question:

We are bringing the 16-bit, 8-band PlanetScope products into GEE to convert them into JPEG tiled images for image segmentation and neural net training - this is why it is important that inter-scene coloring is consistent.

As Rebecca mentioned, Planet’s documentation on converting their imagery to 8-bit (https://developers.planet.com/docs/integrations/gee/gee/) implies that, between scenes, the min and max intensity values for a scene’s R, G, and B bands should be set at a set value to convert to 8-bit, but that is clearly not the case with the products we’re working with. The little preview image that Planet provides in Explorer and in Orders is very different than the result in GEE when using the min and max values Planet suggests. 

I wonder if this has something to do with harmonization with S2, which we used for these data? Does that affect the intensities of the RGB bands and thus throws off the “standardized” min and max intensities for conversion to 8-bit? 

We are just a little frustrated that every scene apparently needs to be tinkered with to get the “right”/best color correction. Any advice from Planet support would be appreciated. 

For that type of application I’d recommend using the PSScene visual bundle, which is a true-color (RGB) rendering of each and every scene that is more-or-less consistent across time and space. https://developers.planet.com/docs/integrations/gee/delivery/

Thanks for sharing this!