Pokémon TCG card pricer

A computer-vision pipeline that recognizes Pokémon trading cards from images or Whatnot livestream footage and estimates a market price with a confidence range. Paired with a six-episode vertical video series documenting the build.

Problem

Prices on the secondary market move fast and depend on set, rarity, and condition. During a livestream there’s no time to look anything up — you want an instant read with an honest sense of how sure the estimate is.

Approach

• Detect and identify the card from a frame (image or livestream capture).
• Map the identification to a price estimate, reported as a range rather than a single number so the uncertainty is visible.
• Document the whole thing as a 6-episode vertical video series.

TODO(JP): name the specific models/techniques you used for detection and identification, and how you built the price lookup.

What broke / what I learned

TODO(JP): livestream frames are messy — glare, motion blur, partial cards. Add the concrete problems and how you handled them.

Next

TODO(JP): next milestone (accuracy target, latency, or publishing the series).