Neural scan AR-Repair-Assistant

"The hands-free AR co-pilot for seamless assembly and maintenance."

Comment

Inspiration

i always like to get to know about my stuff i like to know the details of my hardware my components and also few times when my pc dosent work of some stuff i would like o know about the details etc i used to search on yt google lens the image and long tutorials which was very hectic to see each thing each component one tutorial says remove this other say that and the there are connection wiring or slots on my cpus motherboard i dint knew about and messed up sometimes also figuring out which kind of cell or power source my gadget needed so all these lead to this ai repair components identifier .Hardware repair can be intimidating due to "component anxiety." We built NeuralScan AI to bridge the gap between complex hardware and the everyday user, turning a camera into an expert diagnostic tool that demystifies circuit boards instantly.

What it does

Live Identification: Labels hardware parts (PCIe slots, CPU sockets, etc.) using real-time AR overlays.Health Checks: Automatically assesses component status (e.g., "Healthy" or "Issues Detected"). Technical Insights: Provides a comprehensive summary of the detected hardware architecture and chipset.

How we built it

We utilized a modern web stack for high-performance visual reasoning: Frontend: Built with Next.js and Tailwind CSS for a low-latency, glassmorphic HUD. Intelligence: Integrated Gemini 1.5 Flash for its multimodal spatial grounding capabilities. Spatial Mapping: We translated AI-detected coordinates (0–1000 scale) into screen pixels using:

                          x_{pos} = ({xmin}/{1000}) . width_{screen}
                          y_{pos} = ({ymin}/{1000}) . height_{screen}

Challenges we ran into

Component Differentiation: distinguishing between visually similar components (like different types of capacitors or specific IC chips) was difficult.

Prompt Engineering: getting the AI to strictly adhere to the "Simple" vs. "Pro" user settings was tricky; initially, it would hallucinate technical details in simple mode.

User Experience: We struggled to make the output readable on a small screen while overlaying information on complex circuit boards.

Accomplishments that we're proud of

We are incredibly proud of the Adaptive Explanation Engine. Seeing the AI successfully identify a niche component and explain it simply to a beginner, and then technically to a pro, felt like a massive achievement. We are also proud of the "Hands-Free" implementation, as it solves a genuine physical struggle of repairing hardware (having no hands free to type).

What we learned

We discovered the power of Spatial Intelligence in Multimodal LLMs. Learning to manage high-resolution video streams alongside asynchronous AI inference was key to achieving a fluid user experience.## What's next for AR-Repair-AssistantPredictive Maintenance: Identifying missing or loose connections before they cause failure.Hardware Expansion: Extending support to automotive and industrial machinery.Collaborative AR: Allowing remote experts to "draw" on the user’s live feed.

Future of AR-AI helper

Our building hasn't stopped yet theres so much more ahead for our ai we will train on large datasets real time live tracking and analyzing every component we will build a community forum for people to interact with each other ask post doubts question for help with their gadgets from those who know a community for nerds , tech geeks, people repairing their own stuff, simple question , helps and all rounded community where everyone find answer tho their question

Built With

Share this project:

Updates