Closed james77777778 closed 1 year ago
Is it still open? Can I work on the issue?
Hey @james77777778 thanks for opening this bug.
Some of these are true-positives, and some are false positives. We intentionally do NOT have a namespace in keras_cv.layers.object_detection
, and the same should be made to be true for keras_cv.layers.preprocessing
- these are NOT meant for end users and are instead only supposed to be organizational. The user facing namespace is only keras_cv.layers
.
So with that in mind, all of the layers should be present in keras_cv.layers
!
Closing for now! Thanks for the issue report @james77777778 - the exports will become clearer very soon.
We're working on integrating w/ https://github.com/fchollet/namex to generate our API!
I have noticed that there are some missing imports in
__init__.py
when I work on KPL vectorizing.I use following script to roughly check imports in
keras_cv/layers/
and the outputs:
I'm not sure the missing imports in
keras_cv/layers/object_detection/
is intended or not. Butrandom_aspect_ratio
andjittered_resize
should be included inkeras_cv/layers/preprocessing/__init__.py
I can open a PR once approved.
P.S. This script only works for one class per file pattern. It is not optimal in general but should be effective in
keras_cv/layers/
.