LMVD

We propose a large-scale multimodal video log database (LMVD) for identifying depression in the wild. In LMVD, there were 1823 samples, capturing 214 hours of 1475 participants from four multimedia platforms (Sina Weibo, Bilibili, Tiktok, and YouTube). For all collected data, we extract video features and audio features separately. For audio features, use a pre trained VGGish41 model. For visual features, use FAU, facial markers, eye gaze, and head posture features. It is worth mentioning that our LMVD is the largest dataset for identifying visual and auditory depression in an individual's daily life, which is a positive contribution to the field of emotional computing.