A Python CLI for syncing your Flickr library and deduplicating images using AI or perceptual hash.

• Sync your Flickr photo library to a local folder (skips already-downloaded images)

• AI-powered deduplication of local images:

  • CNN (MobileNetV3): Fast, robust detection of visually similar/near-duplicate photos
  • ONNX CLIP: (Optional) For semantic/visual dedupe using CLIP Vision Transformer

• Text and metadata-based duplicate scans (by title, filename, or date)

• Fuzzy matching for titles

• Designed for performance with large libraries (thousands of photos)

• Python 3.8 or newer
• Flickr API key & secret (get yours here)

Install dependencies:

pip install -r requirements.txt

1. Set your API credentials (in your shell)**:

   export FLICKR_API_KEY=your_api_key
   export FLICKR_API_SECRET=your_api_secret

2. (First time) Authenticate via CLI:

   python cli.py auth

python cli.py sync_flickr --directory /path/to/myphotos --max-images 5000

• Downloads all (or up to --max-images) Flickr photos to the directory, skipping images that are already present.

python cli.py ai_dedupe --directory /path/to/myphotos --method cnn --max-images 1000 --similarity-threshold 0.97

• --method cnn (default): Fast, robust deep-learning dedupe (MobileNetV3)
• --method onnx: (Optional) Uses ONNX CLIP Vision Transformer (requires model)
• Results: prints pairs of duplicates with similarity scores

python cli.py scan
python cli.py fuzzy_scan --threshold 90

• Finds duplicates by title, filename, or date
• Fuzzy scan uses RapidFuzz for inexact title matches

• Use sync_flickr as your main fetcher—run dedupe many times offline!
• For existing archives, skip sync and just point dedupe at your folder.
• The CLI skips already-downloaded images, so you can resume sync at any time.

• Add custom dedupe actions (move, delete, tag)
• Visualize duplicate clusters (see imagededup docs)
• Integrate recursive search or batch reporting

MIT