CASE STUDY

Developed an AI Driven Image Viewer Application for Defect Detection in Buildings using Drone Images

Client

The client is a globally recognized engineering consulting firm that excels in structural engineering, sustainability, protective design, and advanced analytics, offering innovative solutions through cutting-edge technology, including AI and digital twins, to create safe, efficient, and sustainable projects across diverse industries.

Need

The objective was to develop a deep learning-based image viewer application that can be used by organizations and structural project owners to analyze, monitor, and maintain damage detections of structures.

Solution

Damage Detection Architecture:

Developed a deep learning architecture for detecting damages and defects on drone captured images, Orthomosaics, and 2D/3D images.

Lifecycle Management UI:

Developed an interactive UI to maintain the lifecycle of the detections for every project.

Custom DenseNet Model:

Developed a custom deep learning model using DenseNet to improve defect detection results.

Training & Prediction Framework:

Built a comprehensive framework for training and prediction, enabling the model to be easily retrained for new event types using Supervise.ly, Kubernetes, and Kubeflow for ML training and inference pipelines.

Benefits

  • Image viewer application handles various types (2D, 3D, Panoramic images, Video, etc.) and sizes of images taken from drones in one platform.
  • AI-driven defect detection and annotation.
  • Correction of annotated defect images in the application.
  • Dashboard for the user to view all project details.

Result

  • Reduced time spent on defect detection and project monitoring.
  • Improved project efficiency and accuracy through AI-powered solutions.
  • Enhanced image management and defect annotation capabilities.

Technologies

  • Supervise.ly
  • Kubernetes/Kubeflow
  • Node.js
  • Angular
  • three.js
  • Azure services