EMONET-FACE
收藏资源简介:
EMONET-FACE是一个由专家注释的合成情绪识别基准数据集,包含一个新颖的40类情绪分类法,覆盖了更广泛和更详细的人类情感状态。该数据集由三个大规模的AI生成的数据集组成,具有明确的、完整的面部表情和跨种族、年龄和性别的控制人口平衡。EMONET-FACE BIG包含超过203,000张图像,EMONET-FACE BINARY包含近20,000张图像,EMONET-FACE HQ包含2,500张图像。这些数据集通过专家进行严格的注释,以确保高质量的训练和评估。此外,还构建了EMPATHICINSIGHT-FACE模型,在EMONET-FACE HQ基准上实现了人类专家水平的性能。该数据集旨在为开发和评估具有更深层次理解人类情感的AI系统提供坚实的基础。
EMONET-FACE is an expert-annotated synthetic emotion recognition benchmark dataset that features a novel 40-category emotion taxonomy, covering a broader and more detailed spectrum of human emotional states. This dataset comprises three large-scale AI-generated datasets with well-defined and comprehensive facial expressions, as well as controlled demographic balance across race, age, and gender. EMONET-FACE BIG contains over 203,000 images, EMONET-FACE BINARY contains nearly 20,000 images, and EMONET-FACE HQ contains 2,500 images. All these datasets undergo rigorous expert annotation to ensure high-quality training and evaluation. Additionally, the EMPATHICINSIGHT-FACE model was developed, achieving human-expert-level performance on the EMONET-FACE HQ benchmark. This dataset aims to provide a solid foundation for developing and evaluating AI systems with deeper understanding of human emotions.




