Annotation services

Oriented bounding box (rotated box) annotation services

Oriented bounding box (OBB) annotation draws a rotated rectangle that matches each object's angle, stored as four corners or a center, size and rotation. It fits angled objects — ships, parked cars from above, text lines — far more tightly than an axis-aligned box, and trains rotated object detectors such as YOLO-OBB.

Trains
Oriented object detection
From
$0.057 / label
Review
Every job
Formats
YOLO OBB TXT

Is it the right label?

When to use oriented boxes

Oriented object detection — the model finds each object and its heading in a new image.

  • Imagery is top-down or aerial: satellite, drone, overhead cameras, conveyor belts.

  • Objects are elongated and sit at arbitrary angles: ships, aircraft, vehicles in a lot, pallets.

  • Objects are packed together, so axis-aligned boxes would overlap heavily and confuse the model.

When something else fits better

For ground-level photos of upright objects — people, storefronts, traffic from a dashcam — rotation adds labeling time without helping. Use standard bounding boxes.

Bounding boxes

How we label it

Our process

  1. 01

    Agree the heading rule

    Which edge is the front, and whether angle matters beyond fit (ship heading does; a crate's usually does not).

  2. 02

    Fit, then rotate

    Annotators fit the box to the object's long axis so all four sides touch its edges.

  3. 03

    Dense-scene discipline

    In crowded aerial frames every instance is labeled, including partial objects at tile edges, per the guideline.

  4. 04

    Angle and fit review

    Reviewers check fit and heading consistency across similar objects before approving.

Quality

What our QA checks

Every job is reviewed by a person other than the annotator, on top of automated checks. Work below the agreed threshold goes back for rework.

Rotated IoU against reviewer-corrected boxes

Heading consistency on directional classes

Missed small objects in dense aerial tiles

Corner-order and angle-range validation on export

Pricing & delivery

Rates and formats

AnnotationRate
Object detection$0.057 / label

QA review is included. Bulk discounts apply to large projects; the final quote is confirmed after we review a sample of your data. Full pricing

Delivered as

YOLO OBB TXTDOTA TXTCOCO JSON (rotated)CVAT XMLDatumaro

Used for

Satellite and aerial imagery · Ports and shipping · Parking and fleet monitoring · Warehouse pallets from above · Document and text-line detection

Oriented boxes FAQ

Oriented boxes, answered

What is an oriented bounding box?

An oriented bounding box is a rectangle rotated to match the object's angle, described by four corner points or by center, width, height and rotation. It fits angled objects much more tightly than a normal box.

When should I use oriented boxes instead of regular boxes?

Use them for top-down imagery where objects sit at any angle — satellite, drone and overhead camera images — and for elongated objects like ships or text lines. For ground-level photos, regular boxes are usually enough.

Which formats support oriented bounding boxes?

YOLO OBB (four normalised corner points per object) and DOTA are the most common. We also export CVAT XML with rotation and Datumaro.

Are oriented boxes more expensive than regular boxes?

Only slightly per box — the extra step is setting the angle. The bigger cost driver in aerial work is object density, since a single satellite tile can contain hundreds of objects.

Need oriented boxes on your data?

Send us a few sample images or clips. We'll agree the guideline with you and send a plan and a per-label quote.