annotations file? about detect-waste HOT 4 CLOSED

To run the experiments from beining you have to have annotations per dataset in one file. In the scenario for TACO just combination of train and test files.
The main task of annotations_preprocessing.py or annotations_preprocessing_multi.py is to combine together annotations from few sources and make the classes consistent. Just convert to ours categories.
Keep in mind also that in splitting we have some randomnes, so you can create different train and test subsets.

from detect-waste.

Hi,
thank you for the question.
The annotation file which you mentioned is additional annotations provided by our partner epinote for unlabeled photos from TACO dataset. We did not release the part of annotations alone.
For whole extended TACO annotations with 7 classes look in https://github.com/wimlds-trojmiasto/detect-waste/tree/main/annotations. The photos can be downloaded from TACO repository.
Best,
Sylwia

from detect-waste.

I want to start from scratch. When I run: python3 annotations_preprocessing.py, I need to ignore annotations_epi.json, don't I?
Thank you !

from detect-waste.

Thank you very much

from detect-waste.