-
Notifications
You must be signed in to change notification settings - Fork 6
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Is it able to convert point cloud model to tensorflow Probabilistic Deep Learning model ? #1
Comments
The implementation seems correct, but validation loss is sky-rocketing. I assume model is dealing with too extreme values and overfitting on training dataset. Can you remove the line: And try to work with logits: |
Update model and train again.
|
Hmm this is strange, I checked the original Keras doc which has extreme loss values in validation too, so it will be hard to converge for a Bayesian NN. Can you try flipout layers? They should provide less variance in the training if I am not mistaken. They have the same signatures as Reparameterization layers. Ex: Try changing |
Hi Frightera:
I am doing some test with pointnet to classify point cloud object. And there might be some other unknown object are not listed in training set.
So I tried to convert pointnet model(https://keras.io/examples/vision/pointnet/) to tensorflow probability model by refering your Simple Fully Probabilistic Bayesian CNN, and hope the model can say "I don't know" if target object is not in training model, but the results of the training were less than ideal.
Is it possible to convert point cloud model to tensorflow Probabilistic Deep Learning model ?
The text was updated successfully, but these errors were encountered: