Hey,

you can use any model (pretrained or not) with the fasterRCNN model from pytorch. Take a look at the function get_fasterRCNN in fasterrcnn.py in my repo. Just specify your model with the arg backbone. You also might need to change the normalization values.