DeepSignBridge: Sign Language Translator Using Transformers and Machine Vision
Description
Join us on an exploratory journey behind the scenes of “DeepSignBridge”, a pioneering system that translates Peruvian sign language into text in real time. This talk will take you from the foundations of our project, starting with the exploration of NLP architectures such as LSTM and GRU, through the innovative 1-D CNNs, and culminating in the choice of Transformers, which revolutionized our approach. We will delve into the challenges and innovative solutions in pose detection, highlighting the use of cutting-edge tools such as MediaPipe and YOLO Pose, which allowed us to accurately capture complex sign language gestures. In addition, we will share our experiences comparing state-of-the-art models such as ViT and ConvNeXt, and how we finally decided on MaxViT due to its exceptional performance and accuracy. In addition, we will learn how the ChatGPT API can help us improve translation by making it more natural. This talk will not only show you the technology behind DeepSignBridge, but also the impact that artificial intelligence can have in creating a more inclusive world. Discover how perseverance, innovation and technology come together to bridge the gap of inclusive communication.