Making a person who does not understand sign language might be difficult for people who cannot talk or speak appropriately. Priyanjali Gupta, a Delhi-based engineering student, has designed an AI model that can comprehend gestures in order to make their life easier.
Priyanjali is a student at Vellore Institute of Technology in Tamil Nadu, where she is pursuing her engineering degree. Priyanjali’s mother works at Spice Jet and her father is a lecturer at Delhi University.
Priyanjali told a station that her mother inspired her to become an engineer and develop for the greater good. “Nowadays, people are doing so many companies, doing innovation,” her mother informed her as she arrived at her Delhi home. “You are also an engineering student, you should do something.”
Priyanjali, 20, said she’d observed the Alexa device, which responds to voice instructions. She realized that Alexa is useless to individuals who were unable to talk or hear. “That’s when the idea of creating an AI model came to me, and I got to work,” she explained.
“Made this model using this awesome Tensorflow object detection api,” Priyanjali wrote on LinkedIn. “The model uses transfer learning from the pretrained ssd mobilenet model to translate a few ASL signs to English,” she explained.
Priyanjali began developing the concept in February 2021. In December of last year, she created a model based on her computer vision knowledge, but it didn’t turn out properly.
In January, she tried again, this time with the assistance of the internet. ‘I did all the research and hard work myself to create this,’ she explained. I didn’t get any support, but the internet was quite helpful. To get this model to the final stage, I stayed awake for three nights in a row.”
Currently, the Priyanjali model can recognize six signs: hello, I love you, thank you, please, and thank you. Yes, and no. Because the project is still in its early phases, it will undoubtedly take some time to complete. She presently requires a stronger platform and guidance in order to improve the model.