News & Updates

Organic traffic from Google keeps growing, and someone just bought an extremely rare Disney pin for $5,200 on eBay through a link on our site.

Uploads got simpler on mobile and desktop, while the catalog picked up the SEO basics needed to make 155,000 plus pin pages discoverable.

Our biggest model upgrade yet. We trained the v4 pin detection model on over 154,000 images using an NVIDIA RTX PRO 6000 Blackwell GPU with 96GB of VRAM, achieving a 95.3% detection accuracy.

We've built and trained a custom machine learning model right here on our local hardware to automatically detect and identify Disney pins from photos of your boards and lanyards.

A major architecture shift replaced cloud vision with local YOLO detection and CLIP search, making uploads faster and eliminating per-image API costs.

Pin Portfolio now uses a local vision model to score how closely eBay listings match your pin, improving result quality without adding cloud costs.

The pricing pipeline now uses image-aware eBay search, smarter filtering, and caching to return faster and more accurate results.

A huge import brought more than 155,000 Disney pins into a searchable reference catalog, complete with dedicated pages and AI-assisted enrichment.

Pin Portfolio made the jump from local project to live site with Docker deployment, async processing, and better support for iPhone uploads.

Pin Portfolio grew from a solo tool into a community-friendly site with accounts, profile pages, public collections, and a discovery feed.

The first version of Pin Portfolio could take a photo, identify a Disney pin with AI, and pull sold eBay prices to start building a collection.