Viktor.comAn AI coworker that actually does the work
Promoted
PokeScope
Pokemon card price checker & recent eBay sales
50 followers
Pokemon card price checker & recent eBay sales
50 followers
Get instant Pokemon card prices & graded prices, track your collection, and make smarter trading decisions with PokeScope - the ultimate Pokemon TCG companion app.
I created PokeScope.app to make it effortless for Pokémon card collectors to look up prices, graded prices, and see recent eBay sales all for free.
Unlike the generic “AI wrappers” flooding the market, PokeScope uses a custom trained computer vision model that runs locally on your iPhone, built exclusively for Pokémon cards. Our AI scanner (available only in the iOS app) identifies over 96% of cards instantly, giving collectors unmatched accuracy, privacy, and speed.
📊 What makes PokeScope different:
• Custom AI model (not generic OCR)
• Real time eBay sales data (not outdated listings)
• 20,000+ cards in database
• 50k+ collectors trust us (4.8★ App Store rating)
• PokeScope.app website. Free to use, forever
Try it: https://pokescope.app
iOS: https://apps.apple.com/us/app/po...
Very cool app, can you tell us more about the card recognition? That seems to have required a lot of work, what's the tech behind it?
@icobg123 Thanks! Yeah, that part took a ton of work.
The app uses a mix of AI object detection and visual similarity tech to recognize cards in real time. Basically, it looks at your camera feed, finds the card using an object detection model that I trained on thousands of cards, and extracts the card image very accurately, then compares it to 10s of thousands of known cards to figure out exactly which one it is... even down to the set or variant.
It’s all powered by two custom AI models I trained from scratch that run locally on the device, so it’s super fast and works even in tricky lighting, angles, and even when the cards are in sleeves or binders.
@greg_ellis Thanks a lot for the details, that sounds like a ton of work! Do you mind sharing where you got the training data and how you went about training the models? I'm working on MTG app that could really benefit from a similar feature for card recognition.
@icobg123 I found open source pokemon card data sources and then added thousands of my own images that I captured that included cards in sleeves, binders, and different lighting etc. I then augmented the entire data set to apply transformations to the existing images such as rotation, flipping, blurring, etc... This expanded the dataset without me having to take more photos and made the object detection model much more reliable across all kinds of real world conditions.
It's super important for the object detection model to be able to detect the area that the card resides in as accurately as possible so that you can extract the card image with as little of the background image surrounding it as possible. That way when you use the card identification model it will be more accurate and faster especially if you are doing realtime detection like I do in PokeScope.
It’s tedious and painstaking work, but the success of any AI model mostly depends on the size and quality of its training data.
Hope this helps.