Frédéric Branchaud-Charron
Frédéric Branchaud-Charron
Hello, We have a Keras community SIG this Friday (June 5th). I'll raise your issue there.
Yes I will report back on this issue.
For your particular issue: `class_names` is still an attribute (as can be seen [here](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/keras/preprocessing/image_dataset.py#L205)). You can get all the information from [dataset_utils.index_directory](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/keras/preprocessing/image_dataset.py#L175). I know this is not ideal.
From the Keras SIG: With Keras preprocessing Layers coming out of experimental, there will be a lot of documentation on how to transition from the old API to the new....
This PR requires new unit tests.
I think this is a bug, PRs are welcome.
Following this [RFC](https://github.com/keras-team/governance/blob/master/rfcs/20190729-keras-preprocessing-redesign.md), I think we should make Keras preprocessing layers for this task. (Inside TF Repo)
I would stick to fchollet answer and that we only keep it for historical reason. We could deprecate it and remove it. Thoughts @fchollet ?
Add this https://github.com/rykov8/ssd_keras/pull/63/files#diff-53099e21b216bdb89802158e8eaae9baR119 in Keras 2
Actually, the only thing to do is to add the `def compute_output_shape(self, input_shape):` in ssd_layers.