Now that we have trained the model in the usual way, let's look at how we can convert this model into a TensorFlow Lite format.
The general procedure for conversion is illustrated in the following diagram:
The procedure is simple: we take a trained model, freeze the graph, optimize it for inference/prediction, and convert it into .tflite format. Before going further, let's understand what we mean by Freeze Graph and Optimize For Inference:
- Freeze Graph : Freeze graph operation effectively freezes the weights of the model by converting all the of the TF Variables as Constants. As you can imagine, ...
