How to create model with desired outputs on tflite

TensorFlow General Discussion
, ,
1

I have been experiencing with tflite for around 1 month and I learnt a lot with the help of this platform. I would like to learn more specific approaches and now my next issue is below:

I have a script which provides great results for a specific aim with a specific tflite model. The model has 2 outputs which are “output” and “locator” on MobileNetv3 architecture. I would like to create my own model and have tried each architecture on tflite model maker but any of them provides the same outputs so I couldn’t adapt them into my own project. My question is if it is possible to create tflite with that specific outputs or not.

Could anyone provide information on this subject?

Thanks in advance.

3

At the initial phase you can start with prebuilt models in Modelmaker, then you can explore the tflite models with your own architectures(including different custom layers, custom ops and even on different platforms). It is possible to create a TFLite model with 2 specific outputs like “output” and “locator” under MobileNetv3 architecture or with your own architecture. It depends on how the original model you are building and how those outputs are generated.