Train a Mask R-CNN for Instance Segmentation with TF Model Garden

This hands-on tutorial shows how to fine-tune a Mask R-CNN with a MobileNetV2 backbone using TensorFlow Model Garden. You’ll download LVIS annotations, generate TFRecords from COCO images, load a registered Mask R-CNN experiment config, adapt it for a small custom class set, set training hyperparameters, run distributed training, evaluate, and finally export a SavedModel for deployment. Code and configs follow TensorFlow best practices; original samples are CC BY 4.0 / Apache 2.0.