Hi,
I copy the new version of retinanet to another place, but it uses previous version configuration, in the debug.py it says change relative imports to absolute imports to solve the problem. How can I do that?
Relative import :
from ..preprocessing.pascal_voc import PascalVocGenerator
Absolute import:
from keras_retinanet.preprocessing.pascal_voc import PascalVocGenerator
I don't have keras-retinanet instead of that I have detection_net folder, should I use that name?
If you renamed the keras_retinanet folder to detection_net then yes
I add from keras_retinanet_v2.keras_retinanet.models.mobilenet import custom_objects in the first line of my test program, but it throws me below error:
ImportError: No module named keras_retinanet_v2.keras_retinanet.models.mobilenet
Python's imports must match the underlying folder structure, also the keras_retinanet_v2 folder has to be visible to python, so it should be in the Python's path
I don't have keras-retinanet instead of that I have detection_net folder, should I use that name?
Why did you rename it? If you wish to write custom scripts or modules for integrating keras-retinanet with your own application, I would strongly recommend keeping keras-retinanet as it is and writing your own code as an external module that imports keras_retinanet.
Otherwise you will have a very very difficult job keeping up to date with changes from this repository.
Also, if you do wish to fork and rename the repository (you may of course), then support for that is out of scope of this issue tracker.
@smehida is your issue resolved? If so, please close it.
Most helpful comment
Python's imports must match the underlying folder structure, also the keras_retinanet_v2 folder has to be visible to python, so it should be in the Python's path