Add BASNet to keras hub #1984
Add BASNet to keras hub #1984
Conversation
There was a problem hiding this comment.
Started a review. But I think this might need some bigger considerations.
Some questions:
- What are the weight sources we are using here?
- What are the outputs? Looks like there is a list of outputs instead of a normal classification style output.
- How does training this model look?
In general, I think there there will be a lot of core infrastructure that will break if the arch does not have a backbone class. DeepLabV3 is probably much closer to what we will need to go for.
But I think I need answers to the above questions to better suggest a design.
| ) | ||
|
|
||
|
|
||
| @keras_hub_export( |
There was a problem hiding this comment.
you can remove the list for single symbol exports.
@keras_hub_export("keras_hub.models.BASNet")
Also I think in keeping with our other naming this should be "keras_hub.models.BASNetImageSegmenter" right?
| "keras_hub.models.BASNet", | ||
| ] | ||
| ) | ||
| class BASNet(ImageSegmenter): |
| A Keras model implementing the BASNet architecture for semantic | ||
| segmentation. | ||
|
|
||
| References: |
There was a problem hiding this comment.
generally we put these after args I think
| backbone: `keras.Model`. The backbone network for the model that is | ||
| used as a feature extractor for BASNet prediction encoder. Currently | ||
| supported backbones are ResNet18 and ResNet34. Default backbone is | ||
| `keras_cv.models.ResNet34Backbone()`. |
There was a problem hiding this comment.
this is false, update the docstring!
|
|
||
| Example: | ||
| ```python | ||
|
|
| ) | ||
| model.fit(images, labels, epochs=3) | ||
| ``` | ||
| """ # noqa: E501 |
| outputs.extend(predict_model.outputs) | ||
|
|
||
| outputs = [ | ||
| keras.layers.Activation("sigmoid", dtype="float32")(_) |
There was a problem hiding this comment.
generally we use _ as a dummy variable that is unused. this is used, so use x maybe?
| loss=keras.losses.BinaryCrossentropy(from_logits=False), | ||
| metrics=["accuracy"], | ||
| ) | ||
| model.fit(images, labels, epochs=3) |
There was a problem hiding this comment.
How does this work when the model has multiple outputs it looks like?
Refer this issue for more details