I intend to use the Google Open Image Dataset to assist in training an object detection model. Are there certain formats I should use? Are there any instructions to do this?
You can follow our bounding boxes format using the CLI uploader: Uploader
Just be sure to select images from the Google Image subset with bounding boxes (not segmentation). Up to a few hundreds images should work with our free-tier.
So I have been working on this and getting nowhere. I was able to download and install fiftyone and understand how to get subsets of the open image dataset. But after that I am at a loss on how to accumulate them and get them into edge impulse.
If the annotations of your images are in Pascal VOC, you can follow this tutorial to upload your data directly to edge impulse using the edge impulse cli.
Let me know if this might be of help to you.