L-SVD (Large-Scale Selfie Video Dataset (L-SVD): A Benchmark for Emotion Recognition)

Welcome to L-SVD

L-SVD is an extensive and rigorously curated video dataset aimed at transforming the field of emotion recognition. This dataset features more than 20,000 short video clips, each carefully annotated to represent a range of human emotions. L-SVD stands at the intersection of Cognitive Science, Psychology, Computer Science, and Medical Science, providing a unique tool for both research and application in these fields.

Why L-SVD?

Inspired by the revolutionary ImageNet, L-SVD aspires to be a benchmark in the emotional AI space, offering meticulously labeled and consistently processed video data to the global research community.

• Rich Emotional Annotations: Our dataset encompasses a spectrum of eight emotions—Anger, Contempt, Disgust, Enjoyment, Fear, Sadness, Surprise, and Neutral—each annotated with precision.
• Uniform Video Quality: We ensure uniform hue, contrast, and brightness across all videos to provide a standard quality baseline for algorithm development.
• Exponential Growth and Community Driven: L-SVD is rapidly expanding, with new contributions regularly enhancing the depth and breadth of the dataset.

Dataset Features

• Detailed Emotional Range: L-SVD covers a broad array of emotional states, meticulously labeled for accurate recognition.
• Pre-Processed for Excellence: All clips are pre-processed to ensure high fidelity and uniformity, setting a high standard for training data quality.
• Inviting Global Contributions: We encourage contributions from researchers and practitioners around the world to help grow and diversify the dataset.