I consider pre-training a YOLOv5 with Google Open Images Object Detection dataset. The dataset includes general domain categories with ~15 M box samples. After the pre-training is done, I will fine-tune the model on MS COCO dataset.
I would like to do it, if I can improve AP by ~7%. Do you think that it is possible, and I have a logical expectation? Unfortunately, I could not find anywhere anyone has tried an Open Images pre-trained object detector with MSCOCO training.