Iron Solari

Inspiration

Current emergency response systems are often reactive — they rely on victims to identify and report emergencies through a 911 call. This creates delays, especially when victims are unaware of the severity of their situation or are physically unable to call for help. Our team wanted to bridge that gap by creating a system that allows for proactive emergency detection and response coordination, reducing response times and saving lives.

What it does

Iron Solari is an intelligent monitoring platform that uses deep learning and drone imaging to detect emergencies in real time. The system can recognize when people are in distress or when fires are present, triggering instant notifications for emergency responders. Once an incident is detected, our web dashboard updates live to show alerts, recommended actions, and drone locations. Through the interface, users can acknowledge alerts, dispatch first responders, and view the timeline of ongoing or resolved events. The platform also integrates Gemini Nano to analyze and describe images, offering context-specific recommendations for how to respond to each situation. From image capture to response coordination, Iron Solari streamlines every step of the emergency detection process.

How we built it

To bring Iron Solari to life, we began with a DJI Mini 4K drone, which we used to capture aerial photos and videos of test areas. These images are automatically uploaded to Dropbox, pulled locally onto the backend, then transferred to a Supabase database for processing. Our deep learning models analyze each image to detect potential signs of distress or fire. The results are transmitted through websockets to our live monitoring dashboard, built using modern web technologies, where users can view alerts in real time. For the video analysis component, we used FFmpeg to divide footage into five-second clips, which are then analyzed by the model to identify any emergency situations. Once analyzed, Gemini Nano generates natural-language descriptions and recommended responses, enhancing decision-making for dispatchers and users.

Challenges we ran into

One of the biggest challenges we encountered came from the DJI Mini 4K drone itself. Its software ecosystem limited our ability to program low-level logic or use third-party applications to automatically upload images to our database. This restriction made it difficult to create a seamless, automated workflow between the drone and our backend systems. To work around this, we designed a custom macro system that simulated touch inputs, effectively mimicking a user pressing the capture button repeatedly so the drone could continuously take photos. However, because these images weren’t automatically saving to Dropbox, we had to expand the macro to also handle the upload process. This required careful optimization to minimize downtime between captures and uploads, ensuring that our data pipeline stayed as real-time as possible despite the hardware limitations.

Accomplishments that we're proud of

When we first started, Iron Solari could only detect individuals who had fallen over. Over the course of the hackathon, we significantly expanded the capabilities of our machine learning models to identify not just people in distress, but also fires, hazardous situations, and other emergency scenarios. Achieving this level of functionality in under 24 hours was a major accomplishment for our team. It demonstrated our ability to rapidly prototype, iterate, and integrate new AI models into a real-time system under tight time constraints. Seeing Iron Solari evolve from a simple fall detection tool into a comprehensive emergency monitoring platform within a single day was one of the proudest moments of the project.

What we learned

Working on Iron Solari taught us the value of creative problem-solving under pressure. One major roadblock came from the DJI Mini 4K drone, whose restrictive software prevented low-level programming and automated uploads to our database. Initially, this seemed like it could derail the project, but we developed a macro system that simulated repeated touch inputs to capture images continuously and later expanded it to handle uploads with minimal downtime. Overcoming this challenge showed us that even when tools or constraints seem insurmountable, resourcefulness and persistence can keep a complex project on track. This lesson mirrors real-world situations, where emergency responders and engineers alike must adapt quickly to unexpected obstacles to achieve critical results.

What's next for Iron Solari

Looking ahead, we hope to expand Iron Solari’s detection capabilities to include medical emergencies, vehicular accidents, and environmental hazards. We also plan to enable real-time drone navigation so that drones can automatically reposition based on detected activity or alert density. Our long-term goal is to integrate directly with municipal dispatch systems, making Iron Solari a true extension of first responder networks. By improving model accuracy and efficiency with larger datasets, we aim to turn Iron Solari into a proactive guardian system for cities—a tool that ensures faster, smarter, and more reliable emergency responses.

Built With

• nextjs
• node.js
• python
• react
• rest
• socket.io
• supabase
• typescript
• websockets