Memory Map with AI

The Idea

This project integrated the best elements of previous iterations—interactive mapping, real-time image capture, and memory anchoring—with AI-generated descriptions from LLaVA. By automating annotations, the goal was to create spatial narratives enriched with semantic context, simplifying the organization and recall of visual memories.

‍Development‍

Combining Flask for the backend with Leaflet for the front end and OpenCV for image capture, the system featured:

• User-initiated image capture tied to map markers
• LLaVA-driven text captions for each captured image
• Chronologically connected memory points, a fullscreen gallery, and a sidebar for navigation
• A refined dark-themed UI for enhanced user experience

Reflection‍

This iteration achieved a balanced integration of automation and usability. AI-generated captions minimized manual effort, and chronological markers provided a coherent narrative thread. While still lacking AR overlays and facing scalability challenges, the project demonstrated that intelligent, context-aware spatial memory mapping is both feasible and highly promising.

What Worked

• Automated captioning streamlined memory documentation
• Chronological markers offered a narrative-like user journey
• Fullscreen gallery and sidebar improved browsing and management

What Did Not Work

• Scalability concerns for large datasets or continuous image processing
• Lack of on-device personalization reduced adaptability for individual users

‍