HDFD
收藏资源简介:
HDFD数据集是由斯旺西大学计算机科学系创建的,旨在评估非刚性表面注册和分类的挑战性基准数据集。该数据集包含40,000帧的3D模型,来自21名参与者,包括4D有趣面孔和4D视觉-音频面孔,后者是在威尔士语中发音的短语。数据集通过手动标注特征点,并由威尔士专家进一步评估流畅度水平。这些数据对于定量评估注册和分类任务非常有用。此外,数据集还用于评估几种最新的非刚性表面注册技术,展示了其对于未来技术的重要性和实用性。
The HDFD dataset was created by the Department of Computer Science at Swansea University as a challenging benchmark for evaluating non-rigid surface registration and classification tasks. It contains 40,000 frames of 3D models from 21 participants, including 4D expressive faces and 4D audio-visual faces, with the latter being phrases pronounced in Welsh. The dataset features manually annotated facial feature points, and its fluency levels were further evaluated by Welsh domain experts. This resource is highly valuable for the quantitative assessment of registration and classification tasks. Furthermore, it has been utilized to evaluate several state-of-the-art non-rigid surface registration techniques, demonstrating its significance and practical utility for future technological advancements.




