Unverified50% confidenceBenchmarkExact time
在 RK3576 上纯 YOLO 检测模型通常能跑到 30 FPS 以上，而 YOLO26n-Depth 仅有 3-4 FPS
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/18/2026
First Seen
Valid until: 11/16/2026
Sources
Related Claims
UnverifiedYOLO将推理速度提升至45 FPS，Fast YOLO更达155 FPS78% similarUnverified在Jetson Xavier NX上运行TensorRT优化的YOLOv11配合质心追踪，车辆稀少时约27 FPS，大量车辆时帧率下跌到8-11 FPS61% similarUnverifiedYOLOv11对于640×640输入，三个检测尺度（80×80、40×40、20×20）共产生8400个候选框位置59% similarVerifiedYaRN extends the window via frequency-domain interpolation and attention scaling, while LongRoPE achieves more efficient extension by identifying non-uniform redundancy in positional encodings, enabling expansion of the effective context from 4K or 8K to 128K or even million-level tokens57% similarVerifiedThe weights alone of the Llama-3 70B model require approximately 140 GB of VRAM at FP16 precision.57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/770837API
curl https://kongchang.com/api/v1/knowledge/claims/770837MCP
get_claim(id=770837)