0

I'm a newer for Flutter and TensorFlow. I'm developping an face detection app and get face key points.

Learning from https://medium.com/@mundorap2010/face-detection-with-tflite-model-without-firebase-in-flutter-6eadf888f3b0, I don't know how to convert the tflite file to the detection face model. At https://netron.app/ I can found the input and output, but why to create the Anchor and AnchorOption model?

Input and Output

What do type and location represent in input and output?

If I have a new tflite file, I can get the input and output, how to create new face model and use?

I hope to recognize my face through TensorFlow and use my own tflite file, and get the key points of my face.

code shown below:

loadInterPreter() async {
    try {
      interpreter = await Interpreter.fromAsset(
        MODEL_FILE_NAME,
        options: InterpreterOptions(),
      );
    } catch (e) {
      print("Error while creating interpreter: $e");
    }
  }

final outputTensors = _interpreter.getOutputTensors();
for (var tensor in outputTensors) {
  _outputShapes.add(tensor.shape);
}

late final ImageProcessor _imageProcessor = ImageProcessorBuilder()
      .add(ResizeOp(128, 128, ResizeMethod.BILINEAR))
      .add(NormalizeOp(127.5, 127.5))
      .build();

final tensorImage = TensorImage(TfLiteType.float32);
tensorImage.loadImage(image);
final inputImage = getProcessedImage(tensorImage);
TensorBuffer outputFaces = TensorBufferFloat(_outputShapes[0]);
final inputs = <Object>[inputImage.buffer];
final outputs = <int, Object>{
  0: outputFaces.buffer,
};
_interpreter.runForMultipleInputs(inputs, outputs);

but will crash when run to runForMultipleInputs method.

Apach3
  • 41
  • 6
  • I made a very low-level mistake. The input needs 640 * 640, but I wrote 128 * 128. – Apach3 Aug 09 '22 at 08:44
  • All the code is correct, just change this line. `.add(ResizeOp(640, 640, ResizeMethod.BILINEAR))` After getting the output, continue coding according to the output format. So I solved this problem myself, cheers~~ – Apach3 Aug 09 '22 at 08:47

0 Answers0