Image Classification Annotation
for Structured Visual Datasets.
Single-label and multi-label image categorisation with consistent taxonomy application. Pilot phase validates the category definitions before production annotation begins.
What Is Image Classification?
Image classification is the task of assigning one or more category labels to an image from a predefined taxonomy. It is the foundation of many computer vision applications — from content moderation systems that categorise what appears in an image, to medical AI systems that classify scan types, to e-commerce systems that tag products by category.
The quality challenge in classification annotation is taxonomy consistency. When a category definition is ambiguous — when annotators disagree about whether an image belongs to category A or category B — the resulting labels introduce noise into the training signal. This is particularly problematic in multi-label tasks where multiple categories must be applied correctly to the same image.
Aeterna Data addresses this through taxonomy validation during the pilot phase. Before production annotation begins, we annotate a representative sample and measure category-level IAA. Any category with low agreement is reviewed, the definition is clarified, and the guideline is updated before the full dataset is labeled.
Label Types
Single-Label Classification
Each image receives exactly one category label from the taxonomy. The image belongs to one class and only one class. The annotation task is to identify which class is correct.
Examples: Disease classification in medical imaging, scene type classification, product category assignment.
Multi-Label Classification
Each image can receive multiple category labels simultaneously. An image might be labeled as both 'outdoor' and 'vehicle' and 'nighttime' at the same time. Each label is applied independently.
Examples: Content tagging, attribute labeling, multi-condition medical classification.
Use Cases
Medical Imaging
Scan type classification, disease category assignment, and severity level labeling for diagnostic AI training.
Content Moderation
Safe/unsafe content classification, policy category labeling, and severity rating for content moderation systems.
E-commerce
Product category assignment, attribute tagging, and visual search training for retail AI systems.
Document Analysis
Document type classification, form category labeling, and layout type assignment for document AI.
Agriculture
Crop type, disease type, and growth stage classification in aerial and ground-level imagery.
Remote Sensing
Land use category, weather condition, and scene type classification in satellite imagery datasets.
Taxonomy Validation
The most common cause of low-quality classification datasets is an ambiguous taxonomy. Category definitions that seem clear in isolation often produce disagreement when annotators encounter edge cases — images that could plausibly belong to two categories, or images that do not clearly fit any category.
Aeterna Data's pilot phase is specifically designed to surface these problems before they corrupt a full dataset. During the pilot, we annotate a sample that includes representative edge cases, measure inter-annotator agreement at the category level, and identify any category with agreement below threshold.
Edge Case Identification
We identify the image types most likely to cause category disagreement and include them deliberately in the pilot sample.
Category-Level IAA Measurement
Agreement is measured per category, not just overall. A category with low IAA is flagged individually — the rest of the taxonomy proceeds to production.
Guideline Update Before Production
Any ambiguous category receives an updated definition with annotated examples before production annotation begins. The update is documented and applied consistently.
Quality Standard
Every classification 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.
Cohen's Kappa
Pairwise IAA between two annotators on the same image set. Applied on every batch.
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
Project Brief & Scoping
Share your image dataset, category taxonomy, label type (single or multi-label), and any existing annotation guidelines. We scope the pilot — sample size, timeline, and deliverables — and confirm before work begins.
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.
Production Annotation
Full dataset annotation following the validated workflow. IAA measured on every batch. Rework at no cost for any batch below threshold.
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.
How to Start
Send a Project Brief
Share your dataset sample and category taxonomy. If you do not have a taxonomy yet, describe what you need the model to classify and we will advise on category design.
Send BriefReceive a Pilot Proposal
We send a scoped pilot proposal within 48 hours — sample size, timeline, flat-rate pilot fee, and deliverables.
NDA and DPA Signed
Before any data is shared, NDA and Data Processing Agreement are signed. Your dataset stays confidential.
Pilot Begins
Annotation starts inside your platform. IAA report delivered with the pilot dataset. Production follows on your confirmation.
Ready to Classify Your Image Dataset?
Taxonomy validated in the pilot. Category-level IAA measured. Quality report on every batch.