Unverified50% confidenceBenchmarkExact time
MediaPipe Pose can achieve inference speeds of over 30 FPS on mobile GPU via TFLite.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/16/2026
First Seen
Valid until: 11/14/2026
Sources
AI实时格斗计分系统：骨骼追踪网络延迟问题与优化方案
Related Claims
Unverified30–50 TPS is the common speed of many current cloud API services, roughly approximating normal human reading speed.69% similarUnverified将模型权重从FP16压缩到INT8或INT4的量化技术，可使消费级GPU如RTX 4090在本地部署时达到30-80 TPS的实用速度68% similarVerifiedOnline video typically adopts frame rate standards of 30fps or 60fps.64% similarUnverifiedMediaPipe runs in the browser's WebAssembly or WebGL backend using the TFLite (TensorFlow Lite) format, with a model size of only a few MB.64% similarUnverifiedFull-precision video diffusion model inference typically requires a GPU with more than 24 GB of VRAM, whereas 0.4MP with 10-step Turbo acceleration may reduce VRAM requirements to the 12–16 GB level.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/759740API
curl https://kongchang.com/api/v1/knowledge/claims/759740MCP
get_claim(id=759740)