{"id":112825,"date":"2026-10-05T00:57:33","date_gmt":"2026-10-04T16:57:33","guid":{"rendered":"https:\/\/www.wsisp.com\/helps\/112825.html"},"modified":"2026-10-05T00:57:33","modified_gmt":"2026-10-04T16:57:33","slug":"edge-ai%e4%b8%8etinyml%e7%9a%84%e5%88%86%e9%87%8e%ef%bc%9a%e4%bb%8e%e5%b7%a5%e5%85%b7%e9%93%be%e5%b7%ae%e5%bc%82%e5%88%b0%e9%83%a8%e7%bd%b2%e5%86%b3%e7%ad%96%e9%99%b7%e9%98%b1","status":"publish","type":"post","link":"https:\/\/www.wsisp.com\/helps\/112825.html","title":{"rendered":"Edge AI\u4e0eTinyML\u7684\u5206\u91ce\uff1a\u4ece\u5de5\u5177\u94fe\u5dee\u5f02\u5230\u90e8\u7f72\u51b3\u7b56\u9677\u9631"},"content":{"rendered":"

# Edge AI\u4e0eTinyML\u7684\u5206\u91ce：\u4ece\u5de5\u5177\u94fe\u5dee\u5f02\u5230\u90e8\u7f72\u51b3\u7b56\u9677\u9631<\/p>\n<\/p>\n

## 01. \u4e91\u4e2d\u5fc3\u67b6\u6784\u5931\u6548：\u88ab\u7269\u7406\u5b9a\u5f8b\u63d0\u524d\u7ec8\u7ed3\u7684\u5047\u8bbe<\/p>\n<\/p>\n

\u4f20\u7edfIoT\u67b6\u6784\u7684\u903b\u8f91\u5f88\u7b80\u5355：\u4f20\u611f\u5668\u91c7\u96c6\u6570\u636e，\u4e0a\u4f20\u4e91\u7aef，\u5206\u6790\u5b8c\u6bd5，\u8fd4\u56de\u6307\u4ee4\u3002\u8fd9\u5957\u6a21\u578b\u5728\u5de5\u5382\u8f66\u95f4\u3001HVAC\u7cfb\u7edf\u91cc\u7a33\u5b9a\u8fd0\u8f6c\u4e86\u51e0\u5341\u5e74，\u76f4\u5230\u6beb\u79d2\u7ea7\u54cd\u5e94\u7684\u9700\u6c42\u51fa\u73b0\u3002<\/p>\n<\/p>\n

\u4ee5\u89c6\u9891\u76d1\u63a7\u4e3a\u4f8b\u2014\u2014\u6bcf\u8def1080p\u6444\u50cf\u5934\u7801\u7387\u7ea64-8 Mbps，20\u8def\u540c\u65f6\u56de\u4f20\u5c31\u9700\u8981\u81f3\u5c11160 Mbps\u4e0a\u884c\u5e26\u5bbd\u3002\u8fd9\u4e0d\u662f\u94b1\u7684\u95ee\u9898，\u662f\u5149\u7ea4\u7269\u7406\u6781\u9650\u548c\u9aa8\u5e72\u7f51\u6296\u52a8\u51b3\u5b9a\u7684\u53ef\u884c\u8fb9\u754c\u3002\u5f53\u5e94\u7528\u8981\u6c42\u5373\u65f6\u53cd\u5e94\u3001\u79bb\u7ebf\u5bb9\u9519\u3001\u6301\u7eed\u5904\u7406\u97f3\u89c6\u9891\u65f6，\u4e91\u4e2d\u5fc3\u6a21\u578b\u5728\u67b6\u6784\u5c42\u9762\u5c31\u5931\u6548\u4e86\u3002Gartner\u5728\u300aEdge AI Computing Report\u300b\u4e2d\u4e5f\u6307\u51fa，\u52302026\u5e74\u8d85\u8fc750%\u7684\u4f01\u4e1a\u6570\u636e\u5c06\u5728\u6570\u636e\u4e2d\u5fc3\u6216\u4e91\u4e4b\u5916\u751f\u6210\u548c\u5904\u7406\u2014\u2014\u8fd9\u4e0d\u662f\u6982\u5ff5\u7092\u4f5c，\u800c\u662f\u7ea6\u675f\u6761\u4ef6\u5012\u903c\u7684\u5fc5\u7136\u3002<\/p>\n<\/p>\n

## 02. Edge AI\u4e0eTinyML：\u88ab\u6df7\u4e3a\u4e00\u8c08\u7684\u4e24\u79cd\u5de5\u7a0b\u4e16\u754c<\/p>\n<\/p>\n

"Edge AI"\u8fd9\u4e2a\u8bcd\u6b63\u5728\u88ab\u6ee5\u7528\u3002\u5f88\u591a\u4eba\u5c06\u5176\u7b49\u540c\u4e8eTinyML，\u4f46\u4e24\u8005\u7684\u786c\u4ef6\u7ea6\u675f\u548c\u5de5\u5177\u94fe\u903b\u8f91\u5f7b\u5e95\u4e0d\u540c\u3002<\/p>\n<\/p>\n

**TinyML**\u8fd0\u884c\u5728MCU\u4e0a\u3002\u4ee5STM32\u5bb6\u65cf\u4e3a\u4f8b，Cortex-M\u5185\u6838\u8dd1\u572880-240 MHz，SRAM\u4ece\u51e0\u5341KB\u52301 MB\u3002\u6a21\u578b\u5fc5\u987b\u91cf\u5316\u52308-bit\u751a\u81f31-bit，\u4f7f\u7528TensorFlow Lite for Microcontrollers\u2014\u2014\u80fd\u529b\u8fb9\u754c\u662f\u632f\u52a8\u6a21\u5f0f\u8bc6\u522b\u3001\u5173\u952e\u8bcd\u5524\u9192\u3001\u5f02\u5e38\u6e29\u5ea6\u68c0\u6d4b\u8fd9\u7c7b\u8f7b\u91cf\u6a21\u5f0f\u5339\u914d\u3002\u90e8\u7f72\u5de5\u5177\u94fe\u7684\u6838\u5fc3\u662f\u56fa\u4ef6\u70e7\u5f55\u3001JTAG\u8c03\u5f0f\u3001\u4e32\u53e3log\u3002OTA\u6d41\u7a0b\u662f\u56fa\u4ef6\u7b7e\u540d\u3001UDP\u5e7f\u64ad\u3001\u53cc\u5907\u4efdFlash\u5199\u5165\u3001\u91cd\u542f\u5207\u6362，\u8fd9\u5957\u57fa\u7840\u8bbe\u65bd\u7531Arm\u3001\u591a\u5bb6RTOS\u5382\u5546\u548c\u82af\u7247\u539f\u5382\u5206\u522b\u628a\u6301\u3002<\/p>\n<\/p>\n

**Edge AI**\u662f\u53e6\u4e00\u4e2a\u7269\u79cd\u3002\u5b83\u8fd0\u884c\u5728\u5d4c\u5165\u5f0fLinux\u8bbe\u5907\u4e0a，\u6bd4\u5982NVIDIA Jetson\u3001Raspberry Pi CM4，\u751a\u81f3\u5de5\u4e1a\u7ea7x86\u670d\u52a1\u5668\u3002\u7b97\u529b\u8db3\u4ee5\u652f\u6491\u591a\u8def\u89c6\u9891\u6d41\u76ee\u6807\u68c0\u6d4b\u3001\u8bed\u4e49\u5206\u5272\u3001\u4e43\u81f3\u5c0f\u53c2\u6570LLM\u63a8\u7406\u3002\u5de5\u7a0b\u6311\u6218\u662f\u8fd0\u884c\u65f6\u6548\u7387\u3001\u591a\u6a21\u578b\u7f16\u6392\u3001GPU\/NPU\u9a71\u52a8\u6808\u7684\u7a33\u5b9a\u6027\u3002\u90e8\u7f72\u65b9\u5f0f\u662f\u5bb9\u5668\u3001\u955c\u50cf\u3001\u6eda\u52a8\u66f4\u65b0，\u5f00\u53d1\u548c\u8fd0\u7ef4\u7531\u540e\u7aef\u5de5\u7a0b\u5e08\u4e3b\u5bfc\u3002<\/p>\n<\/p>\n

\u4e24\u8005\u7684\u5dee\u5f02\u4e00\u53e5\u8bdd\u6982\u62ec：TinyML\u5728\u6beb\u74e6\u7ea7\u529f\u7387\u4e0b\u68c0\u6d4b\u5f02\u5e38\u632f\u52a8，Edge AI\u670d\u52a1\u5668\u540c\u65f6\u5206\u679020\u8def\u89c6\u9891\u6d41\u3002\u540c\u4e3a"\u9760\u8fd1\u6570\u636e\u6e90"，\u5de5\u5177\u94fe\u3001\u4eba\u624d\u6a21\u578b\u3001\u8fed\u4ee3\u8282\u594f\u5b8c\u5168\u4e0d\u540c\u3002<\/p>\n<\/p>\n

## 03. \u4ee3\u7801\u5bf9\u6bd4：\u4e24\u4e2a\u4e16\u754c\u7684\u63a8\u7406\u5f62\u6001<\/p>\n<\/p>\n

\u8ba9\u6211\u7528\u4e24\u6bb5\u5b9e\u9645\u90e8\u7f72\u8fc7\u7684\u4ee3\u7801\u8bf4\u660e\u5dee\u5f02\u3002<\/p>\n<\/p>\n

### \u573a\u666f\u4e00：TinyML\u632f\u52a8\u5f02\u5e38\u68c0\u6d4b，Cortex-M4，80 MHz<\/p>\n<\/p>\n

\u4ee5\u4e0b\u662fSTM32L4\u4e0a\u8fd0\u884c\u7684TinyML\u63a8\u7406\u903b\u8f91，\u4f7f\u7528TensorFlow Lite for Microcontrollers 2.16.0：<\/p>\n<\/p>\n

```c<\/p>\n

\/\/ TinyML anomaly detection on STM32L4<\/p>\n

#include "tensorflow\/lite\/micro\/all_ops_resolver.h"<\/p>\n

#include "tensorflow\/lite\/micro\/micro_interpreter.h"<\/p>\n<\/p>\n

\/\/ Load model (quantized, 32KB flash)<\/p>\n

const unsigned char* model_data = g_vibration_model_data;<\/p>\n

static tflite::MicroMutableOpResolver<10> resolver;<\/p>\n

static tflite::MicroInterpreter* interpreter;<\/p>\n<\/p>\n

\/\/ 5ms inference loop on 80 MHz MCU<\/p>\n

void infer_vibration(const float* accel_data) {<\/p>\n

\u00a0 \/\/ Copy 240 samples (80ms window @ 3kHz) to input tensor<\/p>\n

\u00a0 memcpy(interpreter->input(0)->data.f, accel_data, 240*sizeof(float));<\/p>\n

\u00a0\u00a0<\/p>\n

\u00a0 \/\/ Invoke interpreter<\/p>\n

\u00a0 interpreter->Invoke();<\/p>\n

\u00a0\u00a0<\/p>\n

\u00a0 \/\/ Read output: failure_probability = 0.84<\/p>\n

\u00a0 float failure_probability = interpreter->output(0)->data.f[0];<\/p>\n

\u00a0\u00a0<\/p>\n

\u00a0 if (failure_probability > 0.80f) {<\/p>\n

\u00a0 \u00a0 \/\/ Local decision, no cloud round-trip<\/p>\n

\u00a0 \u00a0 set_motor_flag(MOTOR_FAULT, true);<\/p>\n

\u00a0 \u00a0 \/\/ Only send meta-data upstream<\/p>\n

\u00a0 \u00a0 upload_telemetry("motor_fault_conf=0.84");<\/p>\n

\u00a0 }<\/p>\n

}<\/p>\n

```<\/p>\n<\/p>\n

\u90e8\u7f72\u8fd9\u4e2a\u6a21\u578b\u65f6\u6700\u6df1\u7684\u4f53\u4f1a\u662f：**\u8c03\u8bd5\u624b\u6bb5\u6781\u5ea6\u53d7\u9650**\u3002\u6ca1\u6709gdb，\u6ca1\u6709printf\u91cd\u5b9a\u5411，\u53ea\u80fd\u9760LED\u95ea\u70c1\u9891\u7387\u548c\u4e32\u53e3\u6ce2\u5f62\u7684\u7ec4\u5408\u6765\u63a8\u65ad\u6a21\u578b\u884c\u4e3a\u3002\u800c\u4e14TFLite Micro\u89e3\u91ca\u5668\u4f1a\u5360\u7528\u7ea620 KB RAM，\u572864 KB\u8bbe\u5907\u4e0a\u8fd9\u5e38\u5e38\u662f\u4e2a\u9884\u7b97\u5669\u68a6\u3002\u91cf\u5316\u540e\u7684\u6a21\u578b\u5982\u679c\u67d0\u4e2a\u5c42\u4e0d\u88ab`MicroMutableOpResolver`\u6ce8\u518c，\u63a8\u7406\u76f4\u63a5\u5d29\u6e83\u2014\u2014\u8fd9\u4e2a\u95ee\u9898\u5728\u7eafPC\u7aef\u5f00\u53d1\u65f6\u5b8c\u5168\u65e0\u6cd5\u9884\u6f14\u3002<\/p>\n<\/p>\n

### \u573a\u666f\u4e8c：Edge AI\u89c6\u9891\u6d41\u5e76\u53d1\u5206\u6790，NVIDIA Jetson Orin NX 16GB<\/p>\n<\/p>\n

\u53e6\u4e00\u4e2a\u6781\u7aef\u662fJetson Orin NX 16GB（\u5185\u5b58\u5e26\u5bbd102 GB\/s，NVIDIA\u5b98\u65b9datasheet\u6570\u636e）\u4e0a\u768420\u8def\u89c6\u9891\u6d41\u76ee\u6807\u68c0\u6d4b，\u63a8\u7406\u5f15\u64ce\u4e3aTensorRT 8.6：<\/p>\n<\/p>\n

```python<\/p>\n

import pycuda.autoinit<\/p>\n

import tensorrt as trt<\/p>\n

import numpy as np<\/p>\n

from concurrent.futures import ThreadPoolExecutor<\/p>\n<\/p>\n

# Load FP16 TensorRT engine<\/p>\n

engine = load_trt_engine("yolov8m_20ch_fp16.trt")<\/p>\n

context = engine.create_execution_context()<\/p>\n<\/p>\n

# Query actual binding shapes from engine<\/p>\n

input_shape = context.get_binding_shape(0) # (20, 3, 640, 640)<\/p>\n<\/p>\n

def infer_stream(camera_id, frame_batch):<\/p>\n

\u00a0 \u00a0 # Bind input\/output buffers for this stream<\/p>\n

\u00a0 \u00a0 d_input = cuda.mem_alloc(trt.volume(input_shape) * 4)<\/p>\n

\u00a0 \u00a0 d_output = cuda.mem_alloc(20 * 25200 * 85 * 4)<\/p>\n

\u00a0 \u00a0 cuda.memcpy_htod(d_input, frame_batch)<\/p>\n

\u00a0 \u00a0 context.execute_async(bindings=[d_input, d_output], stream_handle=stream_handle)<\/p>\n

\u00a0 \u00a0\u00a0<\/p>\n

\u00a0 \u00a0 # Threshold-based filtering on-device<\/p>\n

\u00a0 \u00a0 detections = post_process_nms(d_output, conf_thresh=0.45)<\/p>\n

\u00a0 \u00a0 if any(det.class_id == 0 for det in detections):<\/p>\n

\u00a0 \u00a0 \u00a0 \u00a0 person_detected = True<\/p>\n

\u00a0 \u00a0 \u00a0 \u00a0 # Only push detected-person frames to cloud bucket<\/p>\n

\u00a0 \u00a0 \u00a0 \u00a0 s3_client.upload_cropped_patch(camera_id, detections, frame_batch)<\/p>\n

\u00a0 \u00a0 return person_detected<\/p>\n<\/p>\n

# 20 concurrent video streams with thread pool<\/p>\n

with ThreadPoolExecutor(max_workers=20) as executor:<\/p>\n

\u00a0 \u00a0 futures = [executor.submit(infer_stream, cam_id, frame_queue[cam_id].get())<\/p>\n

\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0for cam_id in range(20)]<\/p>\n

```<\/p>\n<\/p>\n

\u8fd9\u5757\u8bbe\u5907\u5728Orin NX\u4e0a\u7684\u5b9e\u9645\u8868\u73b0\u662f：\u5355\u6b21640×640 YOLOv8m\u63a8\u7406\u7ea69\u6beb\u79d2，20\u8def\u89c6\u9891\u6d41\u6bcf\u8def\u62bd\u5e27\u63a8\u7406，GPU\u5229\u7528\u7387\u7ea670%\u3002\u4f46\u5de5\u7a0b\u4e0a\u7684\u590d\u6742\u5ea6\u8fdc\u8d85\u4ee3\u7801\u5c42\u9762\u2014\u2014TensorRT engine\u9700\u8981\u5728\u76ee\u6807\u8bbe\u5907\u4e0a\u91cd\u65b0\u6784\u5efa，FP16\u7cbe\u5ea6\u5728\u4e0d\u540cbatch_size\u4e0b\u4f1a\u89e6\u53d1\u4e0d\u540c\u7684kernel\u9009\u62e9，\u663e\u5b58\u788e\u7247\u5316\u5728\u957f\u65f6\u95f4\u8fd0\u884c\u65f6\u4e0d\u53ef\u5ffd\u89c6\u3002\u6211\u4eec\u6700\u7ec8\u52a0\u4e86\u4e00\u6bb5\u6bcf6\u5c0f\u65f6\u81ea\u52a8\u91cd\u65b0\u52a0\u8f7dengine\u7684\u903b\u8f91，\u624d\u89e3\u51b3\u4e86\u63a8\u7406\u5ef6\u8fdf\u9010\u65e5\u52a3\u5316\u7684\u95ee\u9898\u3002<\/p>\n<\/p>\n

\u6709\u610f\u601d\u7684\u662f，\u4e24\u6bb5\u4ee3\u7801\u7684\u6838\u5fc3\u903b\u8f91\u9ad8\u5ea6\u76f8\u4f3c：**\u672c\u5730\u63a8\u7406 \u2192 \u9608\u503c\u5224\u65ad \u2192 \u533a\u57df\u4e0a\u62a5**\u3002\u53ea\u662f\u4e00\u4e2a"\u672c\u5730"\u662f\u51e0\u5341KB\u7684RAM，\u53e6\u4e00\u4e2a\u662f\u51e0\u5341GB\u7684\u7edf\u4e00\u5185\u5b58\u3002<\/p>\n<\/p>\n

## 04. \u751f\u6210\u5f0fAI：Edge AI\u8fb9\u754c\u7684\u53c8\u4e00\u6b21\u6269\u5f20<\/p>\n<\/p>\n

\u751f\u6210\u5f0fAI\u6b63\u5728\u6539\u53d8\u8fb9\u7f18\u63a8\u7406\u7684\u5f62\u6001\u3002\u8fc7\u53bb\u8fb9\u7f18\u8bbe\u5907\u5904\u7406\u7ed3\u6784\u5316\u8f93\u5165，\u8f93\u51fa\u6d6e\u70b9\u6570\u6216\u68c0\u6d4b\u6846\u3002\u73b0\u5728\u5c0f\u53c2\u6570\u751f\u6210\u5f0f\u6a21\u578b\u2014\u2014Microsoft Phi-3-mini（38\u4ebf\u53c2\u6570）\u3001Google Gemma-2B\u2014\u2014\u7ecf\u8fc7\u91cf\u5316\u540e，\u57288 GB\u4ee5\u4e0a\u7edf\u4e00\u5185\u5b58\u7684\u8bbe\u5907\u4e0a\u5df2\u7ecf\u53ef\u4ee5\u6d41\u7545\u8fd0\u884c\u3002<\/p>\n<\/p>\n

\u8fd9\u5e26\u6765\u7684\u672c\u8d28\u53d8\u5316\u662f\u4ece"\u5224\u65ad"\u5230"\u751f\u6210"\u3002\u5de1\u68c0\u673a\u5668\u4eba\u4e0d\u518d\u53ea\u8f93\u51fa`failure_probability = 0.84`，\u800c\u662f\u76f4\u63a5\u751f\u6210\u62a5\u544a："\u68c0\u6d4b\u52303\u53f7\u8f74\u627f\u5f02\u5e38\u632f\u52a8，\u6982\u738784%，\u5efa\u8bae\u5728\u672a\u676572\u5c0f\u65f6\u5185\u66f4\u6362\u3002"\u8fd9\u4e2a\u80fd\u529b\u8ba9\u8fb9\u7f18\u8282\u70b9\u4ece\u4f20\u611f\u5668\u9644\u5c5e\u54c1\u53d8\u6210\u5177\u6709\u8bed\u5883\u7406\u89e3\u80fd\u529b\u7684\u81ea\u4e3b\u667a\u80fd\u4f53\u3002<\/p>\n<\/p>\n

\u4f46\u53c2\u6570\u91cf\u8fd8\u662f\u786c\u7ea6\u675f\u3002Phi-3\u91cf\u5316\u52304-bit\u4ecd\u9700\u7ea62 GB\u5185\u5b58，\u8fd9\u8d85\u51fa\u4e86\u7edd\u5927\u591a\u6570MCU\u7684\u80fd\u529b\u8fb9\u754c\u3002\u6240\u4ee5\u751f\u6210\u5f0fAI\u53ea\u80fd\u8fdb\u4e00\u6b65\u62c9\u5927Edge AI\u4e0eTinyML\u7684\u9e3f\u6c9f\u2014\u2014\u524d\u8005\u5728\u5411\u4e0a\u541e\u566c\u4f20\u7edf\u7684\u4e91\u7aef\u63a8\u7406\u8d1f\u8f7d，\u540e\u8005\u4f9d\u7136\u5b88\u7740\u6a21\u5f0f\u5339\u914d\u7684\u8001\u672c\u884c\u3002<\/p>\n<\/p>\n

## 05. \u51b3\u7b56\u9677\u9631：\u786c\u4ef6\u5e73\u53f0\u5373\u5de5\u5177\u94fe\u9501\u5b9a<\/p>\n<\/p>\n

\u9009\u62e9\u786c\u4ef6\u5e73\u53f0，\u5c31\u7b49\u4e8e\u9009\u62e9\u8f6f\u4ef6\u751f\u6001\u3002\u8fd9\u662f\u90e8\u7f72\u51b3\u7b56\u4e2d\u6700\u9690\u853d\u7684\u4ee3\u4ef7\u3002<\/p>\n<\/p>\n

TinyML\u8def\u5f84\u9501\u5b9a\u5728Arm Cortex-M\u751f\u6001：CMSIS-NN\u3001TFLite Micro\u3001\u5404\u5bb6\u7684DSP\u5e93\u3002\u5728\u8fd9\u4e2a\u4e16\u754c\u91cc\u6ca1\u6709Linux\u7684\u5584\u89e3\u4eba\u610f，\u53ea\u6709\u88f8\u673a\u6216RTOS\u7684\u7edd\u5bf9\u63a7\u5236\u3002\u6709\u4e00\u6b21\u9700\u8981\u7ed91000\u53f0\u8bbe\u5907\u505aOTA\u5347\u7ea7，\u56fa\u4ef6280 KB，\u7528\u4e86\u6574\u65744\u5929\u534a\u2014\u2014UDP\u5e7f\u64ad\u7684\u53ef\u9760\u6027\u8fdc\u4f4e\u4e8e\u6709\u4eba\u8109\u7684\u4e91\u670d\u52a1\u3002<\/p>\n<\/p>\n

Edge AI\u8def\u5f84\u5219\u662fLinux\u7684\u5929\u4e0b：Ubuntu 22.04 LTS\u3001Docker CE 24.0\u3001TensorRT\u3001ONNX Runtime 1.17\u3001MLflow\u3002\u90e8\u7f72\u65b9\u5f0f\u4ece"\u70e7\u5f55\u56fa\u4ef6"\u8de8\u8d8a\u5230"\u7f16\u6392\u5bb9\u5668"，\u56e2\u961f\u6784\u6210\u4ece\u5d4c\u5165\u5f0f\u5de5\u7a0b\u5e08\u5207\u6362\u4e3a\u540e\u7aef\u5de5\u7a0b\u5e08\u3002<\/p>\n<\/p>\n

\u8fd9\u4e24\u6761\u8def\u5f84\u4e4b\u95f4\u6ca1\u6709\u5e73\u6ed1\u7684\u659c\u5761，\u53ea\u6709\u60ac\u5d16\u3002\u67b6\u6784\u9009\u578b\u671f\u770b\u4f3c\u65e0\u5173\u7d27\u8981\u7684\u51b3\u5b9a，\u5728\u534a\u5e74\u540e\u90fd\u4f1a\u53d8\u6210\u9ad8\u989d\u8fc1\u79fb\u6210\u672c\u3002<\/p>\n<\/p>\n

## 06. \u667a\u80fd\u843d\u5728\u5408\u7406\u7684\u5730\u65b9<\/p>\n<\/p>\n

\u8fb9\u7f18\u667a\u80fd\u7684\u7ec8\u6781\u547d\u9898\u4e0d\u662f\u8ba9\u4e00\u5207\u8bbe\u5907\u53d8\u667a\u80fd，\u800c\u662f\u5728\u6210\u672c\u3001\u529f\u8017\u3001\u5ef6\u8fdf\u3001\u9690\u79c1\u7684\u7ea6\u675f\u4ea4\u96c6\u5185\u627e\u5230\u6700\u4f18\u951a\u70b9\u3002<\/p>\n<\/p>\n

\u4ece\u90e8\u7f72\u89c6\u89d2\u770b，\u51b3\u7b56\u94fe\u5f88\u6e05\u6670：\u65f6\u5ef6\u8981\u6c4210\u6beb\u79d2\u4ee5\u5185？\u9009Edge\u3002\u6570\u636e\u9690\u79c1\u6cd5\u89c4\u7981\u6b62\u539f\u59cb\u6570\u636e\u51fa\u5883？\u9009Edge\u3002\u8bbe\u5907\u5e38\u5e74\u7535\u6c60\u4f9b\u7535\u4e14\u8ba1\u7b97\u91cf\u4ec5\u4e3a\u6a21\u5f0f\u5339\u914d？\u9009TinyML\u3002\u9700\u8981\u591a\u8def\u89c6\u9891\u6d41\u52a0\u5c0f\u6a21\u578b\u751f\u6210？\u9009Edge AI\u670d\u52a1\u5668\u3002<\/p>\n<\/p>\n

\u968f\u7740\u65b0\u4e00\u4ee3\u5e26NPU\u7684MCU（\u5982NXP i.MX RT1170，\u96c6\u62102MB SRAM）\u8fdb\u5165\u5e02\u573a，TinyML\u6b63\u5728\u7a81\u7834\u8d44\u6e90\u74f6\u9888\u3002\u800c\u624b\u673a\u7aef\u7684Hexagon NPU\u3001Apple Neural Engine\u5219\u8ba9\u6570\u5341\u4ebf\u53c2\u6570\u7684\u7aef\u4fa7AIGC\u6210\u4e3a\u73b0\u5b9e\u3002\u4e24\u4e2a\u65b9\u5411\u90fd\u5728\u81a8\u80c0，\u4f46\u671d\u7740\u4e0d\u540c\u7684\u7a7a\u95f4\u7ef4\u5ea6\u3002<\/p>\n<\/p>\n

"Edge AI vs TinyML"\u7684\u6807\u7b7e\u4e4b\u4e89\u6ca1\u6709\u610f\u4e49\u3002\u771f\u6b63\u7684\u5224\u636e\u662f\u5de5\u4f5c\u91cf\u4e0e\u786c\u4ef6\u7ea6\u675f\u7684\u5339\u914d\u5ea6\u3002AI\u5e94\u7528\u5728\u8fb9\u7f18\u4e16\u754c\u7206\u70b8\u6027\u589e\u957f\u65f6，\u90a3\u4e00\u53f0\u53f0\u5343\u5dee\u4e07\u522b\u7684\u8bbe\u5907，\u6b63\u662f\u667a\u80fd\u5411\u7269\u7406\u4e16\u754c\u6e17\u900f\u7684\u5206\u5f62\u56fe\u8c31\u3002\u6bcf\u4e2a\u90e8\u7f72，\u90fd\u662f\u5bf9"\u667a\u80fd\u5e94\u5728\u4f55\u5904\u8fd0\u884c"\u8fd9\u4e2a\u95ee\u9898\u7684\u4e00\u6b21\u5177\u4f53\u56de\u7b54\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"

# Edge AI\u4e0eTinyML\u7684\u5206\u91ce：\u4ece\u5de5\u5177\u94fe\u5dee\u5f02\u5230\u90e8\u7f72\u51b3\u7b56\u9677\u9631## 01. \u4e91\u4e2d\u5fc3\u67b6\u6784\u5931\u6548：\u88ab\u7269\u7406\u5b9a\u5f8b\u63d0\u524d\u7ec8\u7ed3\u7684\u5047\u8bbe\u4f20\u7edfIoT\u67b6\u6784\u7684\u903b\u8f91\u5f88\u7b80\u5355：\u4f20\u611f\u5668\u91c7\u96c6\u6570\u636e，\u4e0a\u4f20\u4e91\u7aef，\u5206\u6790\u5b8c\u6bd5，\u8fd4\u56de\u6307\u4ee4\u3002\u8fd9\u5957\u6a21\u578b\u5728\u5de5\u5382\u8f66\u95f4\u3001HVAC\u7cfb\u7edf\u91cc\u7a33\u5b9a\u8fd0\u8f6c\u4e86\u51e0\u5341\u5e74，\u76f4\u5230\u6beb\u79d2\u7ea7\u54cd\u5e94\u7684\u9700\u6c42\u51fa\u73b0\u3002\u4ee5\u89c6\u9891\u76d1\u63a7\u4e3a\u4f8b\u2014\u2014\u6bcf\u8def1080p\u6444\u50cf\u5934\u7801\u7387\u7ea64-8 Mbps，20\u8def\u540c\u65f6\u56de\u4f20\u5c31\u9700\u8981\u81f3\u5c11160 M<\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[1198,50,87],"topic":[],"class_list":["post-112825","post","type-post","status-publish","format-standard","hentry","category-server","tag-edge","tag-50","tag-87"],"yoast_head":"\nEdge AI\u4e0eTinyML\u7684\u5206\u91ce\uff1a\u4ece\u5de5\u5177\u94fe\u5dee\u5f02\u5230\u90e8\u7f72\u51b3\u7b56\u9677\u9631 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.wsisp.com\/helps\/112825.html\" \/>\n<meta property=\"og:locale\" content=\"zh_CN\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Edge AI\u4e0eTinyML\u7684\u5206\u91ce\uff1a\u4ece\u5de5\u5177\u94fe\u5dee\u5f02\u5230\u90e8\u7f72\u51b3\u7b56\u9677\u9631 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"og:description\" content=\"# Edge AI\u4e0eTinyML\u7684\u5206\u91ce：\u4ece\u5de5\u5177\u94fe\u5dee\u5f02\u5230\u90e8\u7f72\u51b3\u7b56\u9677\u9631## 01. \u4e91\u4e2d\u5fc3\u67b6\u6784\u5931\u6548：\u88ab\u7269\u7406\u5b9a\u5f8b\u63d0\u524d\u7ec8\u7ed3\u7684\u5047\u8bbe\u4f20\u7edfIoT\u67b6\u6784\u7684\u903b\u8f91\u5f88\u7b80\u5355：\u4f20\u611f\u5668\u91c7\u96c6\u6570\u636e，\u4e0a\u4f20\u4e91\u7aef，\u5206\u6790\u5b8c\u6bd5，\u8fd4\u56de\u6307\u4ee4\u3002\u8fd9\u5957\u6a21\u578b\u5728\u5de5\u5382\u8f66\u95f4\u3001HVAC\u7cfb\u7edf\u91cc\u7a33\u5b9a\u8fd0\u8f6c\u4e86\u51e0\u5341\u5e74，\u76f4\u5230\u6beb\u79d2\u7ea7\u54cd\u5e94\u7684\u9700\u6c42\u51fa\u73b0\u3002\u4ee5\u89c6\u9891\u76d1\u63a7\u4e3a\u4f8b\u2014\u2014\u6bcf\u8def1080p\u6444\u50cf\u5934\u7801\u7387\u7ea64-8 Mbps，20\u8def\u540c\u65f6\u56de\u4f20\u5c31\u9700\u8981\u81f3\u5c11160 M\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.wsisp.com\/helps\/112825.html\" \/>\n<meta property=\"og:site_name\" content=\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"article:published_time\" content=\"2026-10-04T16:57:33+00:00\" \/>\n<meta name=\"author\" content=\"admin\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u4f5c\u8005\" \/>\n\t<meta name=\"twitter:data1\" content=\"admin\" \/>\n\t<meta name=\"twitter:label2\" content=\"\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4\" \/>\n\t<meta name=\"twitter:data2\" content=\"3 \u5206\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/112825.html\",\"url\":\"https:\/\/www.wsisp.com\/helps\/112825.html\",\"name\":\"Edge AI\u4e0eTinyML\u7684\u5206\u91ce\uff1a\u4ece\u5de5\u5177\u94fe\u5dee\u5f02\u5230\u90e8\u7f72\u51b3\u7b56\u9677\u9631 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"isPartOf\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\"},\"datePublished\":\"2026-10-04T16:57:33+00:00\",\"dateModified\":\"2026-10-04T16:57:33+00:00\",\"author\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\"},\"breadcrumb\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/112825.html#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/www.wsisp.com\/helps\/112825.html\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/112825.html#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/www.wsisp.com\/helps\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Edge AI\u4e0eTinyML\u7684\u5206\u91ce\uff1a\u4ece\u5de5\u5177\u94fe\u5dee\u5f02\u5230\u90e8\u7f72\u51b3\u7b56\u9677\u9631\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\",\"url\":\"https:\/\/www.wsisp.com\/helps\/\",\"name\":\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"description\":\"\u9999\u6e2f\u670d\u52a1\u5668_\u9999\u6e2f\u4e91\u670d\u52a1\u5668\u8d44\u8baf_\u670d\u52a1\u5668\u5e2e\u52a9\u6587\u6863_\u670d\u52a1\u5668\u6559\u7a0b\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/www.wsisp.com\/helps\/?s={search_term_string}\"},\"query-input\":\"required name=search_term_string\"}],\"inLanguage\":\"zh-Hans\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\",\"name\":\"admin\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery\",\"contentUrl\":\"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery\",\"caption\":\"admin\"},\"sameAs\":[\"http:\/\/wp.wsisp.com\"],\"url\":\"https:\/\/www.wsisp.com\/helps\/author\/admin\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Edge AI\u4e0eTinyML\u7684\u5206\u91ce\uff1a\u4ece\u5de5\u5177\u94fe\u5dee\u5f02\u5230\u90e8\u7f72\u51b3\u7b56\u9677\u9631 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.wsisp.com\/helps\/112825.html","og_locale":"zh_CN","og_type":"article","og_title":"Edge AI\u4e0eTinyML\u7684\u5206\u91ce\uff1a\u4ece\u5de5\u5177\u94fe\u5dee\u5f02\u5230\u90e8\u7f72\u51b3\u7b56\u9677\u9631 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","og_description":"# Edge AI\u4e0eTinyML\u7684\u5206\u91ce：\u4ece\u5de5\u5177\u94fe\u5dee\u5f02\u5230\u90e8\u7f72\u51b3\u7b56\u9677\u9631## 01. \u4e91\u4e2d\u5fc3\u67b6\u6784\u5931\u6548：\u88ab\u7269\u7406\u5b9a\u5f8b\u63d0\u524d\u7ec8\u7ed3\u7684\u5047\u8bbe\u4f20\u7edfIoT\u67b6\u6784\u7684\u903b\u8f91\u5f88\u7b80\u5355：\u4f20\u611f\u5668\u91c7\u96c6\u6570\u636e，\u4e0a\u4f20\u4e91\u7aef，\u5206\u6790\u5b8c\u6bd5，\u8fd4\u56de\u6307\u4ee4\u3002\u8fd9\u5957\u6a21\u578b\u5728\u5de5\u5382\u8f66\u95f4\u3001HVAC\u7cfb\u7edf\u91cc\u7a33\u5b9a\u8fd0\u8f6c\u4e86\u51e0\u5341\u5e74，\u76f4\u5230\u6beb\u79d2\u7ea7\u54cd\u5e94\u7684\u9700\u6c42\u51fa\u73b0\u3002\u4ee5\u89c6\u9891\u76d1\u63a7\u4e3a\u4f8b\u2014\u2014\u6bcf\u8def1080p\u6444\u50cf\u5934\u7801\u7387\u7ea64-8 Mbps，20\u8def\u540c\u65f6\u56de\u4f20\u5c31\u9700\u8981\u81f3\u5c11160 M","og_url":"https:\/\/www.wsisp.com\/helps\/112825.html","og_site_name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","article_published_time":"2026-10-04T16:57:33+00:00","author":"admin","twitter_card":"summary_large_image","twitter_misc":{"\u4f5c\u8005":"admin","\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4":"3 \u5206"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/www.wsisp.com\/helps\/112825.html","url":"https:\/\/www.wsisp.com\/helps\/112825.html","name":"Edge AI\u4e0eTinyML\u7684\u5206\u91ce\uff1a\u4ece\u5de5\u5177\u94fe\u5dee\u5f02\u5230\u90e8\u7f72\u51b3\u7b56\u9677\u9631 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","isPartOf":{"@id":"https:\/\/www.wsisp.com\/helps\/#website"},"datePublished":"2026-10-04T16:57:33+00:00","dateModified":"2026-10-04T16:57:33+00:00","author":{"@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41"},"breadcrumb":{"@id":"https:\/\/www.wsisp.com\/helps\/112825.html#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.wsisp.com\/helps\/112825.html"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/www.wsisp.com\/helps\/112825.html#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/www.wsisp.com\/helps"},{"@type":"ListItem","position":2,"name":"Edge AI\u4e0eTinyML\u7684\u5206\u91ce\uff1a\u4ece\u5de5\u5177\u94fe\u5dee\u5f02\u5230\u90e8\u7f72\u51b3\u7b56\u9677\u9631"}]},{"@type":"WebSite","@id":"https:\/\/www.wsisp.com\/helps\/#website","url":"https:\/\/www.wsisp.com\/helps\/","name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","description":"\u9999\u6e2f\u670d\u52a1\u5668_\u9999\u6e2f\u4e91\u670d\u52a1\u5668\u8d44\u8baf_\u670d\u52a1\u5668\u5e2e\u52a9\u6587\u6863_\u670d\u52a1\u5668\u6559\u7a0b","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.wsisp.com\/helps\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"zh-Hans"},{"@type":"Person","@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41","name":"admin","image":{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/image\/","url":"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery","contentUrl":"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery","caption":"admin"},"sameAs":["http:\/\/wp.wsisp.com"],"url":"https:\/\/www.wsisp.com\/helps\/author\/admin"}]}},"_links":{"self":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/112825","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/comments?post=112825"}],"version-history":[{"count":0,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/112825\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media?parent=112825"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/categories?post=112825"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/tags?post=112825"},{"taxonomy":"topic","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/topic?post=112825"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}