Depression Detection Using Visual Cues

  • Aditya Shah, Saurabh Mota, Ashit Panchal


Depression is one of the very common mental disorders in the world. Many times, people suffer from depression without knowing. Hence there is a need for automated depression detection. We propose a system which serves as a decision-making backing system for professionals, based solely on features that are extracted from facial expressions and features by interpretation of visual cues of depression. It will predict the scales of the Beck Depression Inventory-II (BDI-II) from visual expressions. Deep learning architectures have a very high accuracy in image recognition and classification. Hence, we propose using deep learning techniques for obtaining better accuracy. Convolutional Neural Network (CNN) is a technique of deep learning that is used for image classification. We propose to use CNN for automated depression detection. The proposed method will be tested on the “Audio/Visual Emotion Challenges 2014” (AVEC2014) dataset.