This is a two-part question:
-
I have a dataset that I would like work with for both object detection and multiclass classification problems. Will I have to upload the same set of images twice, or is there a way to version the dataset such that they can be used for both types of projects?
-
For this dataset the object detection annotations and multiclass labels will be provided as Tensor Flow Object Detection CSV and Multiclass Classification CSV files, respectively. I believe that if I include the annotations as part of the initial upload that the annotations will be available in RF. However, what if I update either of the CSV files and want to re-assign all of the annotations for all the images? Can I use the upload API for this? The examples in the documentation seem to assume a 1:1 correspondence between images and annotation files (i.e., annotations are stored as stand-alone files). In this case all of my images will already be uploaded to RF and I just want to over right ALL the annotations with whatever is included with the updated CSV files (i.e., a single CSV file would have annotations for all the images in the dataset).
Thanks for any suggestions!