Can DataFrame be used in TensorFlowLite for Swift iOs?

General Discussion
, , ,
#1

Hello all:

Apologizes for such basic questions but I am stuck and looking for a different perspective:

Here are my questions:

  1. Can I train a .tflite model in Python on my laptop and deploy it in an iphone swift iOS app?

  2. The python example I followed trains and tests the model with input as a DataFrame. It takes one row at a time from a csv file and passes that to the model. Here is the link to the python example: Human Pose Classification with MoveNet and TensorFlow Lite

  3. In swift ios, (link here: https://github.com/tensorflow/examples/tree/master/lite/examples/pose_estimation/ios) the interpreter.copy expects type Data and does not accept DataFrame.

  4. Are there any examples out there or

  5. best practices I need to be using or

  6. Any direction/hints you could offer?

#3

Hi @Dominic_Pritham, You cannot train the .tflite model. You have to train the keras model and convert it to tflite and can deploy it on IOS as well.

For making predictions using tflite file you cannot pass a dataframe. you have to pass a single example at a time. since the batch size will be one by default.

Please refer to this documnet for more tflite examples.

Thank You.