arduino code

easy_mnist.ino

+#include <Chirale_TensorFlowLite.h>
+
+#include "tensorflow/lite/micro/all_ops_resolver.h"
+#include "tensorflow/lite/micro/micro_interpreter.h"
+#include "tensorflow/lite/schema/schema_generated.h"
+
+#include "arduino_model.h"
+#include "mnist_images.h"
+
+const tflite::Model* model = nullptr;
+tflite::MicroInterpreter* interpreter = nullptr;
+TfLiteTensor* input = nullptr;
+TfLiteTensor* output = nullptr;
+
+constexpr int kTensorArenaSize = 20*1024;
+alignas(16) uint8_t tensor_arena[kTensorArenaSize];
+
+void setup() {
+  Serial.begin(9600);
+  while(!Serial);
+
+  model = tflite::GetModel(arduino_model);
+
+  if (model->version() != TFLITE_SCHEMA_VERSION) {
+    Serial.println("Model provided and schema version are not equal!");
+    while(true); 
+  }
+
+  static tflite::AllOpsResolver resolver;
+
+  static tflite::MicroInterpreter static_interpreter(
+      model, resolver, tensor_arena, kTensorArenaSize);
+  interpreter = &static_interpreter;
+
+  TfLiteStatus allocate_status = interpreter->AllocateTensors();
+  if (allocate_status != kTfLiteOk) {
+    Serial.println("AllocateTensors() failed");
+    while(true); 
+  }
+
+  input = interpreter->input(0);
+  output = interpreter->output(0);
+
+  Serial.println("Ready for input data.");
+}
+
+void loop() {
+  
+  if (Serial.available()){
+    String inputValue = Serial.readString();
+
+    memcpy(input->data.uint8, image2, 28*28);
+
+    TfLiteStatus invoke_status = interpreter->Invoke();
+    if (invoke_status != kTfLiteOk) {
+      Serial.println("Invoke failed!");
+      return;
+    }
+
+    int8_t max_value = output->data.uint8[0];
+    int8_t predicted_digit = 0;
+    for (int i = 1; i < 10; i++) {
+      if (output->data.uint8[i] > max_value) {
+        max_value = output->data.uint8[i];
+        predicted_digit = i-1;
+      }
+    }
+
+
+    Serial.print("Predicted digit: ");
+    Serial.println(predicted_digit);
+    Serial.print("Actual digit: ");
+    Serial.println(image2_label);
+
+  }
+  
+
+
+}