Hello @Lingyu and thanks for taking a look. I have uploaded an archive with two files, using the Dropbox file request. One is in the original RAF format (an example photo only), the other is one of my own, converted to DNG.
The S5 pro (and S3 pro) Fuji cameras have a special “S/R” sensor with two photo sites per pixel: one for the normal exposure, one for extra highlight detail. It appears to me that how these files are interpreted is rather wrong.
For example, the original files should have a resolution of 4256 x 2848 pixels, but when opened in Photo AI, I only see an original resolution of 3030 x 2036 pixels. (the same thing happens when using Sharpen AI.)
In the DNG example file, with the oak tree, you’ll notice that the “R” pixels of the sensor, which are used for the highlights, seem to get ignored entirely. The highlight areas behind the branches cannot be recovered in the Topaz-processed file anymore.