manual best photo selection on miniatures
* generate thumbnails of inbox_clustered (in inbox_clustered_mini) to
allow manual rename and refine clustering
* rearrange original inbox_clustered to mimic structure in
inbox_clustered_mini after rearrangement (i.e. sel and other
subdirectories)
- add web UI to merge subsequent clusters
- generate basic sidecar files
- add sidecar with location to cr2 files if location available in photos from phone
- deep learning:
- detect blurry photos
- face detection
- detect documents
- detect Natalia's selfies
- detect photo series (similar shots) (low resolution comparison?)
- learn which is best photo (representative photo)