BNS - Inspectra AI

Transform property inspections with AI. Analyze photos & videos to detect issues and generate instant reports.

Comment

About BNS Inspectra AI

🌟 The Inspiration

BNS Inspectra AI was born from a desire to create positive impact in Panama's real estate market. We noticed a recurring problem: when people purchase a new home, critical defects often remain hidden until after the property has been delivered. By that time, buyers have already committed, making it difficult and costly to address these issues.

Our mission became clear: empower homebuyers with an AI-powered tool that performs an initial property assessment, giving them actionable insights that can be reviewed by professionals before finalizing their purchase.

🛠️ How We Built It

We designed Inspectra AI as a full-stack web application leveraging modern technologies:

Frontend: Vue 3 with TypeScript and Vite provided a reactive, type-safe interface for rapid development. We integrated jsPDF and html2canvas to generate professional inspection reports with visual markers highlighting detected issues.

Backend: Built on Node.js and Express.js, our API orchestrates the entire workflow—handling media uploads, processing video frames with FFmpeg, and communicating with Google's Gemini AI API for multimodal analysis.

AI Integration: Google Gemini's multimodal capabilities became the heart of our solution, enabling us to analyze both images and videos in a single workflow, detecting structural issues, water damage, electrical problems, and more.

💡 What We Learned

Working with Gemini's multimodal features was transformative. The ability to process diverse media types (images and video frames) simultaneously simplified our architecture significantly—what would have required multiple specialized models now works through a unified AI interface.

We also learned valuable lessons about:

  • Efficiently extracting and processing video frames for analysis
  • Designing intuitive interfaces for complex AI outputs
  • Balancing quick prototyping with scalable architecture
  • The importance of clear visual reporting for non-technical users

🚧 Challenges We Faced

Video Processing: Implementing reliable video frame extraction with FFmpeg required careful handling of different video formats and managing server resources efficiently.

AI Prompt Engineering: Crafting prompts that consistently returned structured, actionable insights took multiple iterations. We needed to balance detail with clarity.

Real-time Feedback: Providing users with progress updates during lengthy video analysis without blocking the interface required thoughtful async handling.

Report Generation: Creating PDF reports with accurate visual markers overlaid on images presented formatting challenges, especially ensuring consistency across different image sizes.

🚀 Looking Ahead

Inspectra AI is just the beginning. Our vision extends beyond inspection to include AI-driven space transformation recommendations. Imagine describing your intended use—"Turn this room into a yoga studio"—and receiving design suggestions, renovation estimates, and visual mockups.

There's tremendous potential to evolve this tool into a comprehensive real estate intelligence platform that serves buyers, sellers, and contractors throughout the property lifecycle.

Share this project:

Updates