Advanced Pokémon Detection and Verification

Principal Investigator:
Peter Sorial

Faculty Advisor:
Yulia Kumar

Abstract:
This project introduces a novel two-stage AI framework aimed at enhancing the detection and verification of Pokémon in images. Initially, the framework employs a Transformer-based model for object detection, accurately identifying Pokémon across various poses and lighting conditions. Following identification, a specialized model assesses whether the images are AI-generated or authentic, ensuring the integrity of digital Pokémon collections and aiding copyright protection.

The researchers are in the process of developing an application intended for deployment on Google Cloud. The evaluation of this application will primarily focus on assessing its accuracy, precision, recall, and its ability to distinguish synthetic images from genuine ones.

This advanced framework promises to make significant contributions to Pokémon research, education, and entertainment. By addressing challenges in synthetic image proliferation, it facilitates the engagement with authentic Pokémon content. Furthermore, its potential extends to various domains, including copyright protection and content management.

Overall, this innovative AI framework holds promise not only for Pokémon enthusiasts but also for industries reliant on image authentication. Its implementation signifies a step forward in leveraging AI for ensuring the integrity and authenticity of digital content.