pyimagesort

Sort image & video files, remove duplicates

Download

https://github.com/monospacesoftware/pyimagesort/releases

Features

• Capable of processing jpg, png, mov, avi, mp4, wmv, m4v.
• Sorts files into destination directory, creating sub-directories for year and month.
• Files are named using the oldest timestamp found on the file, for example /sorted/2017/08/20170801-193758-0.jpg
• Uses filesystem and all available EXIF timestamps to attempt to determine the original (oldest) timestamp
• Duplicates images are detected using by fuzzy match, with rotation.
• Images that have been resized, rotated, very slightly changed, or are almost identical, will be considered matching.
• Always keeps the highest resolution version of matching images.
• If 2 images match and have the same resolution, will keep the HDR version.
• Will preserve the EXIF from the oldest image, even if the oldest was lower resolution and not kept.
• Only deletes duplicate images if an exact match was detected. Otherwise images are backed up to a trash directory.
• Ignores directories starting with . (dot).
• Removes directories emptied as a result of sorting.

Requirements

• Python 3.6+
• exiftool

Installation

• Unzip to any directory
• pip install -r requirements.txt

Usage

sort.py destination source [source ...]

• destination is the full path to a directory to sort images into
• source are full path to directories containing images to process

Notes

• A directory called .imagesort/ will be created in your destination directory, use to store the image database.
• Duplicates images will be backed up to .imagesort/trash. You may delete these as desired.
• Upon startup, the database is checked for consistency with the destination directory, and created or fixed based on images this directory contains.
• You should try to avoid modifying the destination directory once images have been sorted.
• You can adjust the fuzzy match threshold by changing SIMILAR_IMAGE_HASH_DIST in ImageSorter.py.
  • Range 0-1024, default 3
  • 0 = Exact match
  • 3 is conservative, you may want to go as high as 5 or 6. The higher you go the more false matches you'll get and risk losing images you may want to keep.

Known Bugs, TODOs

• Automatic rotation to proper orientation
• Video files do not use fuzzy match

Author

Paul Cowan (Monospace Software LLC)