Akhil Nagori
8th Grade, Juan Cabrillo Middle School
Santa Clara, CA
Visionary: AI Glasses for Real-Time Text-to-Audio Transcription To Help Visually Impaired Students
View Poster
Project Background
When visiting his grandparents in India, Akhil saw his visually impaired grandfather struggle to read Braille. “He had originally had eyesight, but had lost it over time, and it was painful to see him spend so many hours struggling to read the papers,” Akhil says. He also learned that less than one percent of educational books are available to visually impaired kids. Akhil told his teammates Evann Sun and Lucas Yen, and they decided to develop a pair of glasses that could translate text to audio.
Tactics and Results
The team wanted to teach an AI model to “read” different fonts and layouts of text and translate them to speech. They started with a dataset of 800 classroom text images from books, worksheets and other class materials, to train the AI on colorful backgrounds and fonts. The team designed a frame, a pair of 3D-printed glasses that would contain a Raspberry Pi, a tiny camera, a battery, and a small pair of speakers. The glasses were programmed to take pictures, extract the text from them, and used an eSpeak library to convert the text to speech, which is played through the speakers mounted on the glasses’ frame. Akhil oversaw the code to run the glasses. The team tested their glasses in high, low, and medium light with tough to read text. Their glasses successfully read the text aloud 92 percent of the time.
Beyond the Project
Akhil is working on his gardening skills. “Not only does it give me the pleasure of taking care of it,” he says, “but also the delicious harvest I receive at the end of the season” is another reward. Akhil also likes robotics and used access to a makerspace as part of the 49ers STEM Leadership Institute to build a gravity-powered car that can go uphill. Akhil would like to be a computer scientist. “Ever since I was introduced to Python and LeetCode, I was always fascinated by the things computers could do so quickly,” he says.
