An Approach for Object Recognition in Videos for Vocabulary Extraction

English is the most common language globally, and it is increasingly important. English has been compiled in most online documents, information, and contents. However, with a considerable vocabulary, learning English is difficult for many people to remember. Therefore, many modern technologies have been proposed to support English learning, such as English learning technology through word-matching games to help children become excited and easily approach English from an early age. In addition, translation tools can help users look up vocabularies, antonyms, synonyms, and examples. This study presents a method to support learning English via object detection in videos, images, or even live-stream videos in real-time using deep learning architectures such as You Look Only Once (YOLO) - one of the finest families of object detection models with state-of-the-art performances. The method to obtain an mAP is 55.6 with 17GFlops. The results are vocabulary, meaning, and making sentences with that. Our method has good accuracy in data of 2786 images belonging to 59 classes.