Real-Time OCR Translation Optimization

Computer Vision, iOS, Optimization
About This Project


We were tasked to transform a new OCR real-time rendering application that translates the words and languages that it sees on printed materials like signs or objects into a viable product for market. This required the translating and rendering processing to be done in real time, without the need for internet connection.


The initial algorithm used to translate and render the results took upwards of 25 minutes. After optimizing the algorithm, it was reduced an order of magnitude to less than 200 milliseconds. The resulting application is quick, easy to use and very successful in the market.