osprey

Sort a folder of wildlife photos (birds, dolphins, whales) by how sharp the animal is. Each photo moves into A_sharp/, B_soft/, C_blurry/ or D_no_animal/ inside the folder, together with its RAW and XMP files.

Install

Needs a Mac with Apple Silicon and uv.

git clone https://github.com/birdingkit/osprey.git
cd osprey
uv sync

The first run downloads the detector and segmenter (about 800 MB).

Use

uv run osprey ~/Pictures/2026-08-22

[1/3] DSC06266.JPG 60.3 → B_soft/
[2/3] DSC07300.JPG 70.2 → A_sharp/
[3/3] DSC07714.JPG 64.1 → A_sharp/
3 photos in 1.8s: 2 A_sharp, 1 B_soft, 0 C_blurry, 0 D_no_animal

Subfolders are searched too, but everything lands in the four folders at the top. Photos already in those folders are left alone, so running again only sorts new photos. A photo whose name is already taken in its target folder is skipped.

Folder	Animal sharpness (0–100)
A_sharp	61+
B_soft	40–61
C_blurry	under 40
D_no_animal	no bird, dolphin or whale found

The letter prefix keeps the folders in best-first order in Finder.

Files sharing the photo's name move with it: DSC0001.JPG, DSC0001.ARW and DSC0001.ARW.xmp stay together.

osprey moves files, it never deletes them. To undo, move the photos back out of the four folders.

How it works

OWL-ViT finds birds, dolphins and whales from text prompts, and SAM 2 outlines the largest one. Sharpness uses the re-blur metric of Crete et al. (2007) on the animal's pixels, so a soft background does not count against the photo.

A photo takes about 0.5 s on an M4 Mac.

Limits

• Only birds, dolphins and whales are looked for. Add more names to ANIMALS in src/osprey/detect.py.
• Very faint or distant animals, such as a pale seabird soaring against grey sky, can land in D_no_animal. Check that folder before deleting anything.
• The sharpness cut-offs were set on a handful of Sony A7 IV frames and may need adjusting for other cameras or subjects.