Unverified60% confidenceFactExact time
MediaPipe Hands uses a two-stage cascaded pipeline: a palm detector based on BlazePalm and a hand landmark regression model.
2
Sources
60%
Confidence
Long-term
Relevance
8/14/2026
First Seen
Sources
Gesture Synth School：用手势演奏音乐的免费练习应用
Related Claims
UnverifiedMediaPipe hand tracking uses a two-stage detection approach: a palm detector first locates the palm region, then a hand landmark model precisely estimates the 3D coordinates of each joint78% similarUnverifiedGoogle MediaPipe Hands 类框架通过机器学习模型检测手部的 21 个关键点来识别手势72% similarUnverifiedMediaPipe 选择检测手掌而非整只手是一个工程决策，因为手掌作为相对刚性物体外观变化小，检测器可以更小、更快、更稳定，并避免手指自遮挡70% similarUnverifiedMediaPipe Hands 的 21 个关键点覆盖手腕、掌根及每根手指的 MCP、PIP、DIP 关节和指尖，模型在移动设备上推理耗时约 8-10 毫秒69% similarUnverified手部姿态估计技术需要系统实时追踪手指的21个关键节点，包括指尖、关节和掌根66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/743963API
curl https://kongchang.com/api/v1/knowledge/claims/743963MCP
get_claim(id=743963)