Train a YOLO model for object detection with oriented bounding boxes¶

This notebook trains a YOLO model for object detection with oriented bounding boxes on the HRSC2016-MS dataset.

The original dataset can be found here.

We re-use the Tables created in create-custom-obb-table.ipynb.

Project setup¶

[ ]:

PROJECT_NAME = "3LC Tutorials - OBBs"
RUN_NAME = "train-yolon-obb-hrsc2016-ms"
RUN_DESCRIPTION = "Train a yolo11n-obb model on HRSC-2016MS"
DATASET_NAME = "HRSC2016-MS"
MODEL_NAME = "yolo11n-obb.pt"
EPOCHS = 10
NUM_WORKERS = 0

Imports¶

[ ]:

%pip install -q 3lc-ultralytics

[ ]:

import tlc
from tlc_ultralytics import YOLO, Settings

Load tables¶

[ ]:

train_table = tlc.Table.from_names("train", DATASET_NAME, PROJECT_NAME)
val_table = tlc.Table.from_names("val", DATASET_NAME, PROJECT_NAME)

Train a model¶

[ ]:

model = YOLO(MODEL_NAME)
settings = Settings(
    project_name=PROJECT_NAME,
    run_name=RUN_NAME,
    run_description=RUN_DESCRIPTION,
    image_embeddings_dim=2,
)

model.train(
    tables={"train": train_table, "val": val_table},
    settings=settings,
    label_column_name="obb",
    epochs=EPOCHS,
    workers=NUM_WORKERS,
)