Rendered Handpose Dataset

该数据集已用于在我们的论文 Learning to Estimate 3D Hand Pose from Single RGB Images 中训练卷积网络。

它包含 41258 个训练样本和 2728 个测试样本。每个样本提供：

- RGB 图像（320x320 像素）

- 深度图（320x320 像素）

- 类别的分割蒙版（320x320 像素）：背景、人物、每个手指三个类别和每个手掌一个类别

- 每只手的 21 个关键点，它们在图像框架中的 uv 坐标、在世界框架中的 xyz 坐标和可见性指示器

- 内在相机矩阵 K

This dataset was used to train convolutional neural networks in our paper titled "Learning to Estimate 3D Hand Pose from Single RGB Images".

It consists of 41,258 training samples and 2,728 test samples. Each sample includes:
- RGB images (320×320 pixels)
- Depth maps (320×320 pixels)
- Semantic segmentation masks (320×320 pixels): covering background, person, three categories for each finger, and one category for each palm
- 21 keypoints for each hand, with their uv coordinates in the image frame, xyz coordinates in the world frame, and visibility indicators
- Intrinsic camera matrix K

搜集汇总

背景与挑战

背景概述

Rendered Handpose Dataset是一个用于3D手部姿势估计的数据集，包含大量带有深度图、分割蒙版和关键点坐标的RGB图像样本，适用于计算机视觉和机器学习研究。