AeternaData
Object Detection

Bounding Box Annotation
for Object Detection Datasets.

High-consistency 2D object labeling for computer vision training pipelines. Class-specific guidelines, structured quality checks, and IAA measurement on every batch.

2D + 3DBox Types
COCO · YOLOOutput Formats
κ ≥ 0.80IAA Standard
Flat RatePilot Entry

What Is Bounding Box Annotation?

Bounding box annotation is the process of drawing rectangular boxes around objects of interest in an image and assigning each box a class label. It is the most widely used annotation type for object detection — the task of training a model to identify where specific objects appear in an image and what category they belong to.

The quality of a bounding box dataset is determined by two things: how accurately each box is drawn around its object, and how consistently the same class label is applied to the same type of object across the entire dataset. Inconsistency in either dimension introduces noise into the training signal and degrades model performance.

Aeterna Data addresses both dimensions through class-specific annotation guidelines developed during the pilot phase, and inter-annotator agreement measurement on every production batch. Every annotator on a project applies the same rules to the same object types — and that consistency is measured, not assumed.

Use Cases

Autonomous Vehicles

Pedestrian, vehicle, cyclist, and road sign detection for self-driving perception systems.

Retail & E-commerce

Product detection on shelves, planogram compliance, and visual inventory systems.

Security & Surveillance

Person and object detection for video analytics and access control systems.

Medical Imaging

Anatomical structure and pathology localization in radiology and diagnostic imaging datasets.

Robotics

Object localization for robotic pick-and-place, manipulation, and navigation systems.

Drone & Aerial

Vehicle, structure, and land use detection in aerial and satellite imagery datasets.

Annotation Types

2D Axis-Aligned Bounding Box

The standard rectangle drawn horizontally and vertically around an object. Used in the majority of object detection tasks. Supported by all major training frameworks and annotation tools.

Rotated Bounding Box (OBB)

An oriented bounding box that follows the angle of the object rather than the image axis. Used for aerial imagery, document analysis, and any task where objects appear at varied angles and axis-aligned boxes would include excessive background.

Multi-Class Labeling

Each bounding box is assigned one class label from a defined taxonomy. Aeterna Data validates the taxonomy during the pilot phase and documents any ambiguous class boundaries before production annotation begins.

Attribute Tagging

Additional attributes assigned to each bounding box beyond the class label — for example, occlusion level, truncation, or object state. Specified in the annotation guidelines before the pilot begins.

Output Formats

Aeterna Data delivers annotated datasets in the format your training pipeline requires. The output format is specified in the project brief before the pilot begins.

COCO JSON

The standard format for object detection and segmentation tasks. Includes image metadata, category definitions, and per-annotation bounding box coordinates in [x, y, width, height] format. Compatible with most major training frameworks.

YOLO TXT

One text file per image containing normalised class and coordinate values. The standard format for YOLO-family models. Simple structure, fast to parse, and directly compatible with Ultralytics training pipelines.

Custom Format

If your pipeline requires a different output format — Pascal VOC XML, TFRecord, or a proprietary schema — this is specified during scoping and delivered as agreed in the SOW.

Quality Standard

Every bounding box annotation batch delivered by Aeterna Data is measured for inter-annotator agreement before delivery. IAA is not a target — it is a threshold. Batches that do not meet the threshold are reworked before the client receives them.

κ ≥ 0.80

Cohen's Kappa

Pairwise IAA between two annotators on the same image set. Applied on every batch.

κ ≥ 0.75

Fleiss' Kappa

Multi-annotator IAA across three or more annotators. Applied on complex multi-class tasks.

Every batch delivery includes a quality report: IAA scores by class, box count, annotator distribution, and any deviation notes. No dataset delivered without the quality documentation.

Our Workflow

01

Project Brief & Scoping

You share your dataset sample, object classes, and output format requirements. We scope the pilot — sample size, timeline, and deliverables — and confirm before work begins.

02

Pilot Phase

We annotate a representative sample of your dataset following the agreed guidelines. IAA is measured. The quality report is delivered. You review before production begins.

03

Production Annotation

Full dataset annotation following the validated workflow. IAA measured on every batch. Rework at no cost for any batch below threshold.

04

Delivery & Documentation

Annotated dataset delivered in your specified format with a complete quality report — IAA scores, class distribution, and annotator notes.

Annotation Tools

Aeterna Data works with the annotation platform your team already uses. We do not require you to adopt a specific tool. If you do not have a platform, we can advise on setup based on your task type and scale.

CVATLabel StudioRoboflowScale AIV7 DarwinSuperviselyCustom Platform

How to Start

Step 01

Send a Project Brief

Share your dataset sample, object classes, approximate image count, and output format. Use the contact form or email [email protected].

Send Brief
Step 02

Receive a Pilot Proposal

We send a scoped pilot proposal within 48 hours — sample size, timeline, flat-rate pilot fee, and deliverables.

Step 03

NDA and DPA Signed

Before any data is shared, NDA and Data Processing Agreement are signed. Your dataset stays confidential.

Step 04

Pilot Begins

Annotation starts inside your platform. IAA report delivered with the pilot dataset. Production follows on your confirmation.

Ready to Label Your Object Detection Dataset?

Start with a flat-rate pilot. One week. A representative sample. A complete IAA quality report on delivery.