L3D-Pose Dataset

The L3D-Pose dataset is an innovative resource designed for the 2D-to-3D pose lifting task, allowing researchers to effectively reconstruct 3D poses from 2D keypoints. Specifically focused on animal pose estimation and action recognition, this dataset includes both 2D and 3D keypoints, along with detailed skeleton annotations and action labels. Moreover, it features annotations for macaques and horses, making it a valuable tool for machine learning models that aim to analyze animal behavior, estimate poses, and create 3D avatars. Because of its versatility, this dataset is perfect for a wide range of tasks, including computer vision research and AI-driven action understanding.

Key Features:

• 2D and 3D Keypoints: It offers high-quality annotations, which ensure precise pose reconstruction.
• Skeleton Annotations: These detailed structures improve the accuracy of 3D modeling.
• Action Labels: Comprehensive labels make it easier to study and recognize animal actions.
• Animal-Focused Design: It is specifically tailored for macaques and horses, providing unique insights for animal studies.
• Versatile Applications: Researchers can use this dataset for pose estimation, action understanding, and developing 3D avatars.