Konver Keras-Model to RKNN

[ugoebel]

Hello,
I have been trying to convert a Tensorflow/Keras model into an RKNN model for some time now. I have the model here:

import numpy as np
from tensorflow import keras
from tensorflow.keras import layers
import tensorflow as tf
import onnx
import tf2onnx


# 1. MNIST-Modell mit CNN in Keras

# Daten laden und vorbereiten
(x_train, y_train), (x_test, y_test) = keras.datasets.mnist.load_data()
x_train = x_train.reshape(-1, 28, 28, 1).astype("float32") / 255
x_test = x_test.reshape(-1, 28, 28, 1).astype("float32") / 255

# Modell definieren
model = keras.Sequential([
    layers.Input(shape=(28, 28, 1)),
    layers.Conv2D(32, kernel_size=(3, 3), activation="relu"),
    layers.MaxPooling2D(pool_size=(2, 2)),
    layers.Conv2D(64, kernel_size=(3, 3), activation="relu"),
    layers.MaxPooling2D(pool_size=(2, 2)),
    layers.Flatten(),
    layers.Dropout(0.5),
    layers.Dense(10, activation="softmax")
])

# Modell kompilieren
model.compile(loss="sparse_categorical_crossentropy", optimizer="adam", metrics=["accuracy"])

# Modell trainieren
model.fit(x_train, y_train, batch_size=128, epochs=5, validation_split=0.1)

# Modell evaluieren
score = model.evaluate(x_test, y_test, verbose=0)
print(f"Test loss: {score[0]}")
print(f"Test accuracy: {score[1]}")

# 2. Speichern des Modells als ONNX

# Modell als SavedModel speichern
tf.saved_model.save(model, "mnist_model")

# Konvertierung zu ONNX
input_signature = [tf.TensorSpec([None, 28, 28, 1], tf.float32, name='input')]
#onnx_model, _ = tf2onnx.convert.from_keras(model)
onnx_model, _ = tf2onnx.convert.from_keras(model, input_signature, opset=13)
onnx.save(onnx_model, "mnist_model.onnx")

I have not been able to do this so far. Could someone please give me an example of how to do this. I have seen many examples using PyTorch. But I am more familiar with Tensorflow and would like to use it.