Advanced Deep Learning with Python

Implementing Mask R-CNN with PyTorch

In this section, we'll use a pretrained PyTorch Mask R-CNN with a ResNet50 backbone for instance segmentation. This example requires PyTorch 1.1.0, torchvision 0.3.0, and OpenCV 3.4.2. This example is very similar to the one we implemented in the Implementing Faster R-CNN with PyTorch section. Because of this, we'll omit some parts of the code to avoid repetition. Let's start:

1. The imports, classes, and image_file are the same as in the Faster R-CNN example.
2. The first difference between the two examples is that we'll load the Mask R-CNN pretrained model:

model = torchvision.models.detection.maskrcnn_resnet50_fpn(pretrained=True)model.eval()

3. We feed the input image to the network and obtain the output ...