Which is the "BEST" deep learning model for "Custom" object detection for images & real time. YOLO v3, v4, v5, EfficientDet?

Whenever I look for object detection model, I find YOLO v3 most of the times and that might be due to the fact that it is the last version created by original authors and also more stable. In 2020, a new author released unofficial version called YOLO v4 and just after 5 days, another author launched YOLO v5. I am confused that if I have to chose one of the models, which one should I chose? EfficientDet is better than YOLO v3 in my opinion but there is very less talk about that. Is there any other model which is better than thsese? I want to know which models should I chose if I’m detecting objects in images and videos respectively. When speed is not an issue, which one is the best model and when speed is an issue like in videos, which one should I use?