手部关键点数据集

1. 简介手部关键点数据集 26K 包含 26,768 张带有关键点注释的手部图像，专为手部检测和关键点估计模型训练设计；注释通过 MediaPipe 库生成，确保高精度与一致性，且数据集兼容 COCO 和 YOLOv8 格式；每张手标注 21 个关键点，涵盖腕部及拇指、食指、中指、无名指、小指（各 4 个点）；数据集图像来源于 3 个公开平台，按相应许可收集使用，整体分为训练和验证两个子集1. Introduction to the 26K Hand Keypoint Dataset: This dataset contains 26,768 hand images with keypoint annotations, specifically designed for training hand detection and keypoint estimation models. Annotations are generated via the MediaPipe library to ensure high accuracy and consistency, and the dataset is compatible with both COCO and YOLOv8 formats. Each hand is annotated with 21 keypoints, covering the wrist and five fingers (thumb, index finger, middle finger, ring finger and little finger), with 4 keypoints per finger. The dataset images are collected from three public platforms in accordance with their respective usage licenses, and the entire dataset is split into two subsets: training and validation.

搜集汇总

背景与挑战

背景概述

手部关键点数据集包含26,768张手部图像，每张图像标注21个关键点，适用于手部检测和关键点估计模型训练，支持COCO和YOLOv8格式。