Pokémon TCG Search Engine: Use AI to Catch Them All

With a new custom image similarity service, we are able to build an image search engine for collectible cards trading.
Michal Lukáč, Ximilar
Michal Lukac
16. June 2021
Card search engine by Ximilar

Did you play any trading card games? As an elementary school student, I remember spending hundreds of hours playing Lord of the Rings TCG with my good friend. It was a fantastic game. Back then the LOTR was in the cinemas and the cards were simply beautiful. I played with a combination of Ents and Nazguls. Other people in our office spent their youth playing Magic The Gathering (yes, their card artworks are amazing as well), and collecting the sports cards with their favorite players. I personally liked the basketball cards. Cards are still loved, played, collected, and traded by geeks, collectors, and sports fans across the world!

In the last year, we have been building a system able to train visual models with numerous applications in image search engines. We already offer visual search services for photo search. But, they are optimized mostly for general and fashion images. In the last decades, we all witnessed the growth of the TCG cards market. However, technologies based on artificial intelligence have not yet been used in this market. The first system for scanning trading cards was released by ebay.com but it is not available for small shops as API. If you want to have a feature for detecting and recognizing trading cards, just contact us. And since it’s a perfect match, we are going to change itwith a card image search.

Contents:


Where can you use image search for cards?

Trading card games can have tens of thousands of images. In principle, building a basic image classifier leads to low precision and is simply not enough for more complicated problems. However, we are able to build a complex similarity system that can recognize, categorize, and find similar cards by a picture. Once we train it properly, it can deal with enormous databases of images it never encountered before. With this system, you can find all the information, such as the year of release, card title, exact value, card set, or whether it already is in someone’s collection, with just a smartphone image of the card. The collectibles are a big business and some cards are really expensive nowadays. Who knows, maybe you have the card of Charizard or Kobe Bryant hidden in your old box in the attic. We can develop the system for you that can automatically analyze bulk of trading cards sent from your customers.

Which TCG cards could visual AI help with?

An image search engine is a great approach when the number of classes for the image classification is high (above 1.000+). With cards, each card is representing a unique class. A convolutional neural network (CNN) trained as a classifier can have poor results when working with a larger number of classes. Pokémon TCG contains more than 10.000 cards (classes), Magic the Gathering (MTG) over 50.000, and the same goes for basketball or any other sports cards. So basically, we can build a visual search system for both:

  • Trading card games (MTG, Lord of the Rings, Pokémon, Yu-gi-oh!, Warhammer, …)
  • Sport collector cards (Topps, Upper Deck, Panini, …) like Ice Hockey, Football, Soccer, Baseball, Basketball, UFC, and more.
Pokémon, Magic The Gathering, LOTR, Ice Hockey and Basketball cards.
Pokémon, Magic The Gathering, LOTR, Ice Hockey, and Basketball cards.
Yes, we are big fans of all these things 🙂

A similar image recognition technology is starting to be used on e-bay platform when listing trading and sports cards for sale. However, this is only available in the e-bay app on smartphones. The app has a built-in scanning tool for cards and can find the average price with additional info. Our service for card image search can be integrated into your website or application. And you can simply connect via API through a smartphone, computer, or sorting machine to find exact cards by photo, saving a lot of time and improving the user experience!

We’ve been recently training a CNN model for Pokémon cards image recognition (トレーディングカード). Why Pokémon? Pokémon is the most played TCG in the world, the game has simple rules and an enormous fan base. Some cards are really expensive, but more importantly, they are traded heavily!

With this model, we built a reverse search for finding the exact Pokémon card, which achieved 94%+ accuracy (i.e. exact image match). And we are still talking about a prototype developed in a few days that can improve to almost 100 %.

Automatic grading and inspection of cards with AI

A lot of companies are grading sport & trading cards manually. Our visual AI can be trained to detect corner types, scratches, surface wear, light staining, creases, focus, borders. The Image recognition models are able to identify marks, miscut, off center, print defects and other special qualifiers. For example PSA is a company that has develop their own grading standards for automatic card grading (MINT). With our platform and team, you can automatize entire workflow of grading with just one photo.

PSA graded baseball card. Our machine learning model can analyze picture of these cards.
PSA graded baseball card. Automatic grading is possible with machine learning.

With the new custom similarity service, we are able to create a custom solution for trading card image search in a matter of weeks. The process for developing it is quite simple:

  1. We will schedule a call and talk about your goals. We will agree on how we will obtain the training data that are necessary to train your custom machine learning model for the search engine.
  2. Our machine learning specialists will assemble a testable image search collection and train a custom machine learning model for you in a matter of weeks.
  3. After meeting all the requirements of PoC, we will deploy the system to production, and you can connect to it via Rest API.

Machine learning models bring endless possibilities not only to pop culture geeks and collectors, but to all fields and industries. From personalized recommendations in custom fashion search engines to automatic detection of slight differences in surface materials, the visual AI gets better and smarter every day, making routine tasks a matter of milliseconds. That is one of the reasons why it is an unlimited resource of curiosity, challenges, and joy for us, including being geeks – professionally :). Any idea came to your mind? Let’s discuss it.

Michal Lukáč, Ximilar

Michal Lukáč ML Engineer & Co-founder

Michal is a co-founder of Ximilar and a machine learning expert focusing mainly on image recognition, visual search and computer vision. He is interested in science, loves reading books and choking people at Brazillian Jiu-Jitsu trainings.

Related Articles

A comprehensive tutorial for converting a .mp4 .mkv or .mov videos to the streaming formats (HLS or DASH) with Python and FFmpeg.
Read moreAugust 2022
Enhance your images' resolution without losing quality with a powerful Image Upscaler based on visual AI.
Read moreMay 2022
With the AI Explainability in Ximilar App, you can see which parts of your images are the most important to your image recognition models.
Read moreDecember 2021