LS3D-W
收藏资源简介:
LS3D-W是由诺丁汉大学计算机视觉实验室创建的当前最大且最具挑战性的3D面部地标数据集,包含约230,000张图像。该数据集通过结合所有现有数据集并使用2D地标引导的CNN转换为3D地标来创建。创建过程涉及使用先进的神经网络架构和训练技术。LS3D-W主要应用于3D面部对齐问题,旨在解决面部识别和分析中的高精度需求,特别是在处理大姿态和复杂表情时的挑战。
LS3D-W is currently the largest and most challenging 3D facial landmark dataset developed by the Computer Vision Laboratory of the University of Nottingham, which contains approximately 230,000 images. This dataset is created by integrating all existing facial landmark datasets and converting them into 3D landmarks via 2D landmark-guided CNNs. Its creation process employs advanced neural network architectures and training techniques. LS3D-W is primarily applied to 3D facial alignment tasks, aiming to address the high-precision requirements in facial recognition and analysis, especially the challenges posed by large head poses and complex facial expressions.
- 1How far are we from solving the 2D & 3D Face Alignment problem? (and a dataset of 230,000 3D facial landmarks)诺丁汉大学计算机视觉实验室 · 2017年



