Hi,
I am willing to export the MaskRCNNPredictor
of
torchvision.models.detection.mask_rcnn
To TensorRT via torch2trt or via onnx. My doubt is if I will have problems with some unsupported layers. For instance, is FrozenBatchNorm2d()
supported? I was looking them here Support Matrix :: NVIDIA Deep Learning TensorRT Documentation but just saw BatchNormalization
Thank you for your time