Accuracy of a tensorflow lite model

TensorFlow General Discussion
, ,
1

I tried to use the custom dataset training (for object detection) using TensorFlow lite model maker. I used the function given in the tutorial link,

*model = object_detector.create(train_data, model_spec=spec,..............*

But after each epochs, I can’t see the accuracy of the training dataset or validation set… How can I see them at each epochs?

3

Hi @codybinz,

I would like to put an update on this backlog issue. Usually object detection models do not display directly the training process accuracy metrics. However you can write a customized code for performance metrics in the training loop itself. Here is the sample code:

    loss, accuracy, precision, recall, f1_score = model.evaluate(
    validation_data, verbose=1  # Print detailed metrics
)
print(f"Epoch {epoch+1}: Loss - {loss:.4f}, Accuracy - {accuracy:.4f}")

Right now, the model_maker is having issue with Python>3.9. So try the model_maker with Python 3.9 with customized evaluation metrics.

Thank you