Research Intern - Deep Learning
03-2017 - 08-2017 IT development Topic: Talking-face Detection using Audio and Visual Modalities in TV Shows. Python, Keras and Scikit-Learn are used. Source code on GitHub: https://github.com/dyabmo/plumcot
○ Trained convolutional neural networks and long short-term memory networks (LSTM) to detect talking faces in videos, using both visual and audio modalities.
○ Accomplishments: Showed that using a combination of visual and audio modalities for LSTM improves the performance over using either modality alone.