a question about this method

[xjyisok]

hello
So your work doesn’t need to train again, and just use some method to utilize the detect results of different frames which are input to the pretrained backbone like retinanet.
Do you think it is possible to change the retinanet with yolo