Curate dataset from COCO 2017. We mix the original train and val together.
-
Split the Data: Programmatically split the curated dataset into training, validation, and test sets (e.g., 80%/10%/10%). Ensure this split is reproducible by using a fixed random seed RANDOM_SEED = 42.
-
Format for Training: Convert the dataset into the format required by the chosen detection framework (Here is YOLOv8n). Document this process and provide any conversion scripts.
