esp32 部署环境和vscode安装，esp32安装

Hello world project deploy in ESP32 Board. # How to Create a hello world project from TF Lite: I follow the instruction in tensorflow guideline to create this project (including components folder, main folder). ## Clone tensorflow v2.2.0 ``` git clone https://github.com/tensorflow/tensorflow.git cd tensorflow ``` Switch to version 2.2.0 which is more stable for this project than newest versions. ``` git checkout v2.2.0 ``` ## Generate then example The examples can be generated by the following command: ``` make -f tensorflow/lite/micro/tools/make/Makefile TARGET=esp generate_hello_world_esp_project ``` ## Install ESP IDF You can follow the instruction to install ESP-IDF to build the project in README_ESP.md. ## Building the example Go the directory of the example you generated before with command: ``` cd tensorflow/lite/micro/tools/make/gen/esp_xtensa-esp32/prj/hello_world/esp-idf ``` To build this, run: ``` idf.py build ``` ## Load and run the example To flash (replace `/dev/ttyUSB0` with the device serial port): ``` idf.py --port /dev/ttyUSB0 flash ``` Monitor the serial output: ``` idf.py --port /dev/ttyUSB0 monitor ``` Use `Ctrl+]` to exit. The previous two commands can be combined: ``` idf.py --port /dev/ttyUSB0 flash monitor ``` # Train a sinusoid model and create a harmonic motion with LEDs ## Train a sinusoid model You can train a sinusoid model by many ways and can easily import the model you trained into the project above. I suggest you to see an example for training a sinusoid model and convert it using [TensorFlow Lite and xxd](https://colab.research.google.com/github/tensorflow/tensorflow/blob/master/tensorflow/lite/micro/examples/hello_world/train/train_hello_world_model.ipynb). ## Import your model into example project Get your model.cc that you have from the previous step that we have convert a quantized model. Rename it into sine_model_data.cc, and paste it in **main** folder. ## Control the LED In this project we would like to deploy our model on ESP32 Board and stimulate the harmonic motion. So I have add some basic logic handlers and control our LED through driver/gpio.h. In main_function.cc, we have 2 majors change to setup GPIOs and control LEDs. ### Setup GPIOs Include some libraries for GPIO and setup some GPIO that we need to use later. ```c #include "freertos/FreeRTOS.h" #include "freertos/task.h" #include "driver/gpio.h" gpio_num_t pinArray[6] = { GPIO_NUM_0, GPIO_NUM_4, GPIO_NUM_16, GPIO_NUM_17, GPIO_NUM_5, GPIO_NUM_18 }; void setupGPIO() { gpio_config_t gpioConfig; gpioConfig.pin_bit_mask = 0; for (int i = 0; i < 6; i++) { gpioConfig.pin_bit_mask |= (1 << pinArray[i]); } gpioConfig.mode = GPIO_MODE_OUTPUT; gpioConfig.pull_up_en = GPIO_PULLUP_DISABLE; gpioConfig.pull_down_en = GPIO_PULLDOWN_DISABLE; gpioConfig.intr_type = GPIO_INTR_DISABLE; gpio_config(&gpioConfig); } ``` Load our models with TF Lite (in setup()). ```c static tflite::MicroErrorReporter micro_error_reporter; error_reporter = &micro_error_reporter; model = tflite::GetModel(g_sine_model_data); if (model->version() != TFLITE_SCHEMA_VERSION) { TF_LITE_REPORT_ERROR(error_reporter, "Model provided is schema version %d not equal " "to supported version %d.", model->version(), TFLITE_SCHEMA_VERSION); return; } // This pulls in all the operation implementations we need. // NOLINTNEXTLINE(runtime-global-variables) static tflite::ops::micro::AllOpsResolver resolver; // Build an interpreter to run the model with. static tflite::MicroInterpreter static_interpreter( model, resolver, tensor_arena, kTensorArenaSize, error_reporter); interpreter = &static_interpreter; // Allocate memory from the tensor_arena for the model's tensors. TfLiteStatus allocate_status = interpreter->AllocateTensors(); if (allocate_status != kTfLiteOk) { TF_LITE_REPORT_ERROR(error_reporter, "AllocateTensors() failed"); return; } // Obtain pointers to the model's input and output tensors. input = interpreter->input(0); output = interpreter->output(0); ``` Basic input, output for inference. ```c input->data.f[0] = x_val; // Run inference, and report any error TfLiteStatus invoke_status = interpreter->Invoke(); if (invoke_status != kTfLiteOk) { TF_LITE_REPORT_ERROR(error_reporter, "Invoke failed on x_val: %f\n", static_cast<double>(x_val)); return; } // Read the predicted y value from the model's output tensor float y_val = output->data.f[0]; ``` Create function to choose LED after running inference with our model, in this circuit board, I use 6 LEDs so I need to split into 6 equal segments in range [-1. 1] ```c int chooseLED(float y) { if (-1 <= y && y < -0.66) return 0; if (-0.66 <= y && y < -0.33) return 1; if (-0.33 <= y && y < 0) return 2; if (0 <= y && y < 0.33) return 3; if (0.33 <= y && y < 0.66) return 4; if (0.66 <= y && y <= 1) return 5; return -1; } ``` Now for each loop(), we can have an inference of a y_val from x_val with our model, and then control our LEDs to blink in a specific amount of time. ```c for(int i = 0; i < 6; i++) { gpio_set_level(pinArray[i], 0); } int iLED = chooseLED(y_val); printf("Choose LED: %d\n", iLED); gpio_set_level(pinArray[iLED], 1); vTaskDelay((1000 / SEGMENT) / portTICK_PERIOD_MS); gpio_set_level(pinArray[iLED], 0); ``` # Create a circuit board for stimulator a harmonic motion with LEDs Here is an example of a circuit board that I designed for this project in tinkercad.com with Arduino UNO.  # Result [](https://youtu.be/6RySA9YDytg)