Leobit
AI-Based Image Analysis Solution
Custom development of an AI-powered image processing tool with object detection, captioning, and tagging functionality.
ABOUT the project
- Client:
- Leobit's Internal Project
- Location:
-
USA
- Company Size:
- 100+ Employees
- Industry:
-
Information Technology
- Solution:
- Custom Software Development
An AI-powered image analysis solution for intelligent image processing. It can automatically extract objects from images, generate image captions, tag objects, and process large volumes of unstructured visual data.
The solution uses the functionality of Azure Computer Vision enhanced with custom code, as well as convenient and interactive user interface built with Angular and TailwindCSS. It holds a significant potential for integration and implementation across different industries, ranging from healthcare to financial services.
The solution holds a significant potential and can be applied for processing unstructured visual data across industries. It facilitates data categorization, helps detect objects or persons in images, and provides features for digitizing documents. The tool and its functionality can be enhnaced to achieve greater precision and make it suitable for various domains.
Customer
It was our internal project aimed at testing the capabilities of Azure Computer Vision and empowering our actual and potential customers.
Business Challenge
Processing large volumes of unstructured visual data from images is a common challenge across various industries, ranging from healthcare to education. The Leobit team decided to build a solution to this problem while exploring the capabilities of Azure Computer Vision.In particular, searching for different types of materials across datasets may be a challenging task due to issues with image categorization. The Leobit team decided to build a solution that responds to the challenge.
Project
in detail
While the project was completed in the short term, it still involved several major stages, namely planning, configuration of Azure AI Vision Image Analysis, and the development of the app powered with AI functionality.
AI-Based Image Analysis
Our solution applies custom Azure AI Vision Image Analysis configurations to process images with high accuracy. It efficiently detects objects, providing coordinates for bounding boxes that are automatically generated with a custom algorithm based on Fabric.js. Azure AI vision Image Analysis also provides capabilities for AI-powered image classification and processing that support efficient and precise image tagging.
The tool also uses Azure AI Vision’s optical character recognition (OCR) capabilities to enable fast and accurate document digitization. Additionally, its algorithms can be fine-tuned to improve precision, efficiently process larger volumes of unstructured visual data, and ensure more accurate object detection using bounding boxes aligned with object contours.
Processing of Large Volumes of Unstructured Data
The solution provides advanced functionality for searching through unstructured data, which allows users to locate particular objects, persons, or animals. These capabilities can enhance various tasks, such as creating presentations, enhancing media content, or streamlining content discovery. Our tool also supports image classification and grouping, such as organizing photos by the individuals depicted.
With such functionality, teams can collect datasets for powering machine learning tasks, building recommendation engines, and developing systems that find similar content. It is worth noting that we at Leobit use the solution internally, evolving from basic data store searches to a more intelligent, data-driven approach.
Simple and Convenient UI
To ensure usability of the solution, we created a custom UI powered with Angular and TailwindCSS for styling. Our specialists used Fabric.js to ensure interactive image rendering with highlighted boundaries of objects identified and marked by Azure AI Vision Image Analysis. To connect the given service with the app’s UI, we used .NET as a back-end technology. All the processed images are temporarily stored in the Azure Blob Storage.
Automated Image Tagging for Detected Objects
The solution automatically generates tags with concise descriptions of detected objects, including their classification and specific properties. In addition, it can generate simple yet accurate descriptive captions. The tool also assigns confidence scores to the objects it manages to identify across images. Therefore, even if a solution fails to identify an object with a 100% confidence, it leaves notes that simplify human human review of an image.
Explore
The solution prototype
Explore a PoC image analysis platform that can be integrated into custom applications across industries, providing visual data processing support for daily operations, compliance, and service quality.
Technology Solutions
- Custom Azure AI Vision Image Analysis configurations for an AI-powered image processing.
- A user-friendly interface that seamlessly integrates with Azure AI Vision, providing bounding boxes and descriptive tags for objects identified on the image, as well as OCR capabilities for capturing text on images.
- Cloud-based architecture based on .NET, Azure AI Vision, and Azure Blob Storage.
Value Delivered
- An efficient tool for processing large volumes of unstructured visual data.
- The tool that can be applied across various industries, ranging from healthcare to education.
- Rich possibilities for continuous optimization and training for an image processing model.
- Optical Character Recognition (OCR) features for efficient digitization of documents and other text on images.
