Brought to you by the Medical Science Center Computer Vision Group at the University of Wisconsin Madison, EmotionNet is an extensive and rigorously curated video dataset aimed at transforming the field of emotion recognition. This dataset features more than 5,000 short video clips, each carefully annotated to represent a range of human emotions. EmotionNet 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.

Inspired by the revolutionary ImageNet, EmotionNet 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: EmotionNet is rapidly expanding, with new contributions regularly enhancing the depth and breadth of the dataset.

• Detailed Emotional Range: EmotionNet 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.

Contributions are vital for EmotionNet to thrive. For instructions on how to contribute, please visit our contribution guidelines.

For information on how to access and use EmotionNet for your research or applications, please consult our usage documentation.

# Example code to load the EmotionNet dataset

import emotionnet

# Load dataset
dataset = emotionnet.load('/path/to/emotionnet')

# Loop through the dataset
for video in dataset:
    frames, emotions = video['frames'], video['emotions']
    # Insert your model training or evaluation code here

If you use EmotionNet in your academic or industry research, please cite it as follows:

@misc{emotionnet2023,
  title={EmotionNet: A Comprehensive Video Dataset for Emotion Recognition},
  author={Peiran L, Linbo T, Xizheng Y. 
Medical Science Center Computer Vision Group, University of Wisconsin Madison},
  year={2023},
  publisher={\url{https://github.com/PeiranLi0930}},
  journal={*},
  howpublished={\url{https://github.com/PeiranLi0930/emotionnet}},
}

EmotionNet is released under the BSD-3-Clause license.

For support or further inquiries, please contact us at [email protected].

We acknowledge the collective efforts of all contributors from the University of Wisconsin Madison's Medical Science Center Computer Vision Group and the global research community. Your insights and contributions are shaping the future of emotion recognition technology.