Overview / Usage

Sign language translation has been a long standing computer vision problem. Almost all the approaches so far require special hardware for effective results. In this approach, hand pose estimation will be used to extract the locations of key points of a hand and then they will be fed as inputs to the encoder in the sequence to sequence model. Using CTC, the relevant hand pose vectors can be matched with corresponding translations. For the entire model to work, a large dataset of sign language is needed.