We ❤️ Open Source
A community education resource
How I use computer vision to organize old family photos
How open source tools like CLIP, YOLOv8, and FiftyOne turned 1,000's of photos into a searchable archive.
This article is part of the eBook: Everyday AI guide: Practical genAI life hacks from real users, a free download from We Love Open Source.
Like many people, I have folders full of old family photos scattered across different devices—phones, hard drives, even scanned albums. The problem? They’re completely unorganized. Want to find that photo of Grandma’s birthday from 2004? Good luck scrolling through thousands of files named IMG_1234.jpg.
That’s where AI and computer vision became my secret life hack.
The everyday photo organization problem
My photo library was a mess: Duplicates, blurry shots, random screenshots, and lots of untagged people and events. Manually organizing everything felt impossible. I needed a way to quickly sort, search, and tag the collection without spending weeks dragging and dropping files.
My step by step workflow
Here’s how I used open-source AI + computer vision tools to turn chaos into order.
Collect & load images
The first step was gathering all my scattered folders and images from phones, old hard drives, scanned albums and pulling them into FiftyOne, a dataset visualization tool. From there, I loaded a Vision Language Model (VLM) to run a zero-shot classification pass. This step was surprisingly powerful: It automatically tagged photos by place, action, or event, and even separated images with people. To make searching smarter, I also ran CLIP to generate embeddings, giving me a way to measure similarity, duplicity, and uniqueness across thousands of images.
Clean duplicates
Once I had embeddings, the next obvious problem was duplicates. Normally, this means scrolling through endless near-identical shots with slightly different smiles, bad angles, or the same photo saved twice. With similarity scores, I could spot and delete duplicates in minutes instead of hours.
Tag people with FiftyOne
With the clutter reduced, it was time to focus on the people that mattered most. FiftyOne’s similarity search let me pick a single photo of someone, say Grandma, and instantly pull up every related image across the dataset. That was the first time I saw all her birthday photos grouped together automatically.
Refine with CVAT + YOLOv8
Of course, tagging faces isn’t perfect out of the box. That’s where CVAT, an annotation tool, came in. I drew bounding boxes around specific faces, then fine-tuned a YOLOv8 model for face recognition using my dataset. It sounds technical, but the payoff was huge: The model learned to consistently identify my family members.
Apply your model at scale
Once YOLOv8 was trained, I ran it across the entire dataset. Suddenly, I could pull up any picture of family members from the 90’s to today in seconds. That felt like magic!
Visualize & filter with FiftyOne
Finally, I went back into FiftyOne to filter by event, person, or tag. Searching for “cake + Grandma” popped up every birthday celebration instantly. Vacations, holidays, and baby photos all became searchable collections without me having to drag and drop a single file.
The result
What used to be a chaotic digital attic turned into a searchable, organized archive of memories. I can now:
- Find photos by person, place, or event without endless scrolling
- Create quick albums (like “all summer vacations” or “all baby photos”)
- Finally, share organized collections with family members
Conclusion
AI often feels distant and reserved, typically associated with tech companies or researchers. But in reality, open source tools make it practical for everyday use. Organizing old photos might sound small, but it’s about preserving memories, reducing stress, and reclaiming time.
If you’ve got a messy photo library (and who doesn’t?), try giving AI a shot. You might be surprised how much clarity it brings to your memories.
Models & tools used:
- Vision-Language Model (for zero-shot classification)
- CLIP (for embeddings & similarity search)
- YOLOv8 (fine-tuned for face recognition)
- CVAT (annotation tool for bounding boxes)
- FiftyOne (for dataset visualization & filtering)
Get the “Everyday AI” eBook
This free guide highlights simple, real-world examples of how people are using AI to save time, spark creativity, and make life easier.
The opinions expressed on this website are those of each author, not of the author's employer or All Things Open/We Love Open Source.