Kadaabra

The Kadaabra project began with a vision to create a mobile app capable of showcasing real-time outfits and accessories worn by actors in films and TV shows. Driven by curiosity and the challenge, I passionately contributed as a co-founder, exploring innovative sound recognition solutions.

Service introduction video

First prototype of the automatic scene/actor recognition software I developed

Responsibilities

lightbulb

Design

I personally designed every technical aspect of the project. This includes infrastructure, protocols, audio recognition, and video processing. Technologies used: AWS, LAMP, Python/Dejavu, Appcelerator Titanium.

dns

Infrastructure

The infrastructure is designed to be scalable and highly available on the AWS platform. Components include LAMP web servers, audio recognition servers, MySQL/Aurora database servers, and backend systems.

videocam

Video Processing

I developed a Python-based software using the OpenCV library and its wrapper to extract a frame for each "scene" of a video along with metadata about the actors present.

api

API

I designed and developed the APIs to facilitate communication between the app and server infrastructure. The comprehensive documentation crafted enabled smooth integration by the app developer.

graphic_eq

Sound Recognition

For sound recognition, I leveraged the MIT-licensed "Python Dejavu" library, optimizing it for this purpose. Originally developed for music recognition, the library was adapted for the narrower frequency spectrum of TV show dialogue.

groups

Coordination

I coordinated the other members of the development team, including a web developer and a mobile developer.

arrow_back Back to Portfolio Visit Website open_in_new