Emo3D
收藏资源简介:
Emo3D数据集由伊朗德黑兰沙里夫理工大学的NLP与DH实验室创建,旨在解决3D面部表情生成中的情感描述问题。该数据集包含150,000个实例,每个实例包括情感文本描述、对应的图像和3D混合形状分数。通过使用GPT-3.5生成情感描述,并利用DALL-E 3生成图像,结合MediaPipe框架提取混合形状分数,Emo3D数据集为训练和评估3D面部表情生成模型提供了丰富的资源。该数据集主要应用于动画设计、虚拟现实和情感人机交互等领域,旨在提高用户体验和真实感。
The Emo3D dataset was developed by the NLP and DH Lab at Sharif University of Technology in Tehran, Iran, to address the challenge of emotional description in 3D facial expression generation. This dataset contains 150,000 instances, each including an emotional text description, a corresponding image, and 3D blend shape scores. The emotional descriptions are generated using GPT-3.5, images are produced with DALL-E 3, and blend shape scores are extracted via the MediaPipe framework. The Emo3D dataset provides a rich resource for training and evaluating 3D facial expression generation models. It is mainly applied in fields such as animation design, virtual reality, and affective human-computer interaction, with the goal of enhancing user experience and realism.

- 1Emo3D: Metric and Benchmarking Dataset for 3D Facial Expression Generation from Emotion Description伊朗德黑兰沙里夫理工大学计算机工程系NLP与DH实验室 · 2024年



