A basic project to generate an instance segmentation dataset from public datasets such as OpenImagesV6 with FiftyOne. And that this dataset generated in YOLOv8 format is used to train a detection and segmentation model with Ultralytics.
In fiftyone/fiftyone.pdf you can find information of how FiftyOne library works to generate datasets. In the example, a dataset with instance segmentation of 'Cats' from OpenImagesV6 is downloaded. Then is exported to YOLOv5 format (only detections). And a function is included to generate also instance segmentation labels in YOLOv8 format. All code is available in fiftyone/data.ipynb
In yolov8/yolov8.pdf you can find information of some of the main functions of Ultralytics library with YOLOv8: inference, train and export model. yolov8/inference.ipynb contains the code to inference a image with YOLOv8 and how to work with the obtained results (bounding bound, mask, category, confidence) yolov8/train.ipynb contains the code to load a pretrained YOLOv8 model, train it with the custom dataset and export it in other formats.