possible feature request

[Bootsektor]

Would it be possible to implement the following features:

• scan more than one folder , encode the found pictures and save the encodings inclusive path/filename - so this time consuming task has to be done only once
• save the list in a format it can be read back on start / or put it into a DB - so the efforts already done are saved
• compare the database against a single picture encoding with given thresholds. This would enable a way faster processing of single pictures or a folder with new pictures. New pictures in the "to process folder" - oder pictures no longer available as they are present in the DB.

As for now and i hope i have understand this correct - the encoding results encodings=method_object.encode_images('foldername') can not be used in find_duplicates_to_remove or find_duplicates. Of course - once the original images are no longer present - an plot will not be possible, but as the hashes are known - new files can be checked against the hashes and marked as duplicates.

converted the picture/hash arrays to json and saved them - reload and reconverted it to ndarray - but how to use it to calculate duplicates and check against new files?