False annotation

Hi, so I am very new to making an object detection model. I’m trying to upload a Stanford cars dataset to use to train it. I uploaded an image and a coco format annotation to annotate an image, but the result is every class that should be car models turned out to be all 1 class called json.

Am I doing anything wrong with my coco JSON file? Did I do it wrong?
My coco json

Hi @BikiniGordon I took a deeper look at the JSON file and I’m not particularly sure, as it looks correct. It could be due to the ordering of the id, supercategory, and name fields within categories.

I looked at getting the dataset and annotation files from Kaggle to debug and test myself, but I wasn’t sure what changes were made to them, as the Kaggle version has the annotation files in .mat format rather than COCO JSON format.

Did you reformat the annotation files yourself? If so, can you create a Google Drive folder of the images you tried to upload too and share it with me mohamed@roboflow.com?

If all images you used for upload, including the file names, are the exact same as the Kaggle ones, with no changes or omissions, let me know, and I can go ahead and download them to test).