CODEWITHSUNDEEP
tensorflowmachine-learningarduinotensorflow-litearduino-nano
ofithch79
ofithch79

Reputation: 195

Printing the contents of the TensorFlow Lite Input Tensor

I'm trying to run a TF Lite model on the Arduino Nano BLE 33 board.

My data comes from two ultrasonic sensors -- Channel 1 and Channel 2. I am trying to determine where an object might be, so the sensors are measuring distance.

Every time I run my inference, my code returns the exact same answer -- that with 100% accuracy there is an object on the left side of the test area. I get this answer ever when there is no object there.

I am taking 5000 samples per channel (this is the samplesRead figure) and was loading them in as follows:

tflInputTensor->data.f[samplesRead*2+0] = channel1;
tflInputTensor->data.f[samplesRead*2+1] = channel2;

Where channel1 and channel2 are the samples from the sensor.

As I said, every time I run my inference I got the exact same answer.

I decided to cut my code back to try and understand what was going on, so, I hard-coded some figures in, as follows:

tflInputTensor->data.f[samplesRead*2+0] = 4512.00;
tflInputTensor->data.f[samplesRead*2+1] = 4428.00;

But I still got the same answer, no matter what figures I put in here. This makes me think I am loading the data in incorrectly, so, I wanted to print the contents of the input tensors to check what was in there. And this is what I am hoping I can get some help with.

Also, I have 2 in tflInputTensor->data.f[samplesRead*2+0], because I seen an example that had 6 inputs, and in this bit of code they had 6, so I just modified that.

For completeness, here is that main section of code:

const int numSamples = 5000;
int samplesRead = numSamples;
void loop() 
{
  float durationCh1, durationCh2;

  while(samplesRead ==  numSamples) 
  {
    samplesRead = 0;
    break;
  }
  // for(i=0; i<5000; i++)  
  while(samplesRead < numSamples)
  {  

    tflInputTensor->data.f[samplesRead*2+0] = 4512.00;
    tflInputTensor->data.f[samplesRead*2+1] = 4428.00;

    samplesRead++;    
    Serial.println(samplesRead);

    if(samplesRead == numSamples)
    { 
 
      TfLiteStatus invokeStatus = tflInterpreter->Invoke();
      Serial.println("Invoke");      
      if (invokeStatus != kTfLiteOk) 
      {
        Serial.println("Invoke failed!");
        while (1);
        return;
      }

      // Loop through the output tensor values from the model
      for (int i = 0; i < NUM_LOC; i++) 
      {
        Serial.print(GRIDS[i]);
        Serial.print(": ");
        Serial.println(tflOutputTensor->data.f[i],9); // The int here gives the decimal places
      }
      Serial.println();
      delay(2000);
    }
  }
}

Upvotes: 0

Views: 187

Answers (1)

ofithch79
ofithch79

Reputation: 195

I was obviously a sleep deprived idiot when I posted this question, because the answer was in my code!!

I modified Serial.println(tflOutputTensor->data.f[i],9); to be:

Serial.println(tflInputTensor->data.f[samplesRead]);

Upvotes: 0

Related Questions