Case Studies
- Supporting Universities and Research Institutes in USA, Denmark and UAE for their AI researches through our Annotation services
- Polygon annotations to AI companies in Sweden, Singapore and USA
- Point cloud / 3D LiDAR annotations to AI companies in USA, Germany and Japan
- How Annotation Support Helped a Recycling Company Improve Waste Sorting Accuracy?
- Annotation support to a Research Institution to improve people mobility
- Annotation services to an Aqua culture farm
Introduction
In partnership with AI companies in Sweden, Singapore and The United States, Annotation Support offered high accuracy polygon annotation services for advanced computer vision applications. The project included annotation of an annotated complex set of images in the context of Object Detection, Instance Segmentation, Autonomous systems, Smart Surveillance and Industry and AI solutions.
The clients held the need for precise object boundaries which were not adequately represented through conventional bounding box annotations.
Challenge
The AI companies struggled with a number of issues:
- Any complex object with nonlinear boundaries involving point-by-point accuracy.
- Tons of image data to be consistently labelled.
- An application that requires a machine learning model with a high level of accuracy during training.
- Short product realization and deployment deadlines.
- Ensuring consistency in annotations across various types of objects.
Some inaccuracies in the annotation have the potential to affect how the model performs and the trustworthiness of AI predictions.
Our Solution
Annotation Support designed a customized polygon annotation workflow to meet client requirements.
Annotation Services Delivered
- Polygon annotation for vehicles, pedestrians, machinery, and infrastructure assets.
- Instance segmentation models: precise boundary identification for objects.
- Multi-class object annotation across a variety of datasets.
- Creating, sharing, and adapting annotation and metadata guidelines for consistency across projects.
- Special quality assurance and validation activities carried out.
Quality Control Measures
- Multi-level review system
- Trained computer vision annotation experts
- Random sampling and accuracy audits
- Continuous feedback integration from client AI teams
Results
The project delivered measurable benefits across all three regions:
- Achieved 95%+ annotation accuracy.
- Successfully performed more than 500K annotations on image objects.
- Cut dataset preparation time by 40%.
- Enhanced segmentation model performance and improved object recognition accuracy.
- Accelerated the implementation of AI models across the organization.
Business Impact
The annotated datasets enabled clients to create more accurate computer vision representations of complex visual environments. They made it easier to get better performance in autonomous systems, industrial automation, smart city and advanced analytics platforms, due to the improved data quality.
The AI companies could concentrate on model development and innovation, knowing that the AI workload of polygon annotation was being handled by an external company, Annotation Support, and that they could rely on a consistent and good-quality supply of training data.
Conclusion
AI companies all over the world still need Annotation Support that will support them in various ways. Today, scalable and qualitative polygon annotation services are still being offered by Annotation Support for AI companies all over the world. Businesses that implement solutions based on our AI and Algorithm and Deep Learning capabilities benefit from our deep expertise in complex image labeling, enhance the accuracy of their models, and quickly launch new computer vision solutions.
