{"slug": "gizmo-guard-safeguard-bot-powered-by-gemma4", "title": "Gizmo Guard - Safeguard Bot (Powered by Gemma4)", "summary": "The Gizmo Guard is a privacy-focused, edge-based AI safety bot that uses a Raspberry Pi and ArduCam to monitor workspaces for scene changes, such as a moved mug. When motion is detected, it captures evidence images and sends them to a Spring Boot backend, which uses a locally running Gemma 4 model for multimodal image reasoning and natural-language explanations. The system is designed to run affordably without cloud infrastructure, ensuring all camera data remains local.", "body_md": "This is a submission for the Gemma 4 Challenge: Build with Gemma 4\nGizmoGuard is a low-budget, privacy-first AI-at-the-edge personal safety and monitoring bot powered by locally running Gemma models.\nThe idea started from a simple but relatable problem:\n“Who moved my mug?”\nGizmoGuard continuously monitors a workspace — or any valuable object of interest, indoors or outdoors — using an ArduCam attached to a Raspberry Pi. The system detects scene changes such as:\nThe system is designed to intelligently distinguish between normal environmental activity and a real scene change near the protected object.\nWhen motion or scene changes are detected, GizmoGuard captures “evidence images” and sends them to a Spring Boot backend API. The backend then uses Gemma 4 for multimodal image reasoning and natural-language explanations.\nUsing additional preconfigured contextual information, the system can also:\nThe entire system is built around a local-first AI architecture:\nGizmoGuard demonstrates how compact multimodal AI models like Gemma can power practical, privacy-focused real-world edge AI applications.\nThe current GizmoGuard architecture consists of the following components:\nThe Spring Boot backend acts as the orchestration layer and:\nPowered by Gemma 4, the AI layer:\nThe project demonstrates how practical multimodal AI systems can run locally using affordable hardware — without requiring expensive cloud infrastructure or hosted AI services.\nDemo Link: Gizmo-Guard Bot Demo\nDemo Includes\nMug placed on desk\nScene continuously monitored by Raspberry Pi + ArduCam\nMug moved, removed, or scene unexpectedly changes\nEvidence image captured automatically\nGemma analyzes the image and explains what changed using multimodal reasoning\nWhen real people (or images of them) appear in the scene:\nGitHub (sasiperi) Repo name and Link: gizmo-guard-gemma4-challenge\nTech stack includes:\nGizmoGuard is powered by Gemma 4B Quantized (gemma4:4B-Q4_K_XL\n) running locally through Docker Model Runner (DMR).\nI specifically selected this model because it delivered the best overall balance between:\nOne of the primary goals of GizmoGuard was ensuring that camera images and personal workspace data never leave the local environment.\nBy running Gemma locally:\nFor an always-on visual monitoring system, this was extremely important.\nI evaluated several local multimodal models.\nSome lightweight models were fast but struggled with:\nLarger models produced strong results but required significantly more resources and slower inference times.\ngemma4:4B-Q4_K_XL\nturned out to be the ideal middle ground:\nThis made it an excellent fit for AI-at-the-edge workloads.\nA major advantage of Gemma4:4B\nwas its ability to handle:\nwithin a single model.\nThis avoided the need to chain together:\nUsing a unified multimodal model simplified:\nAnother goal of the project was proving that useful AI systems do not require expensive cloud GPUs or recurring API fees.\nRunning Gemma locally means:\nThis makes GizmoGuard practical for:\nGizmoGuard demonstrates how compact multimodal models like Gemma can power practical real-world edge AI applications using affordable hardware and open-source tooling.\nThe project combines:\ninto a fully working end-to-end system.\nIt showcases how modern multimodal AI can move beyond cloud-only deployments and become useful directly at the edge.", "url": "https://wpnews.pro/news/gizmo-guard-safeguard-bot-powered-by-gemma4", "canonical_source": "https://dev.to/sasiperi/gizmo-guard-safeguard-bot-powered-by-gemma4-200", "published_at": "2026-05-21 21:21:45+00:00", "updated_at": "2026-05-21 22:04:47.182051+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "large-language-models", "open-source", "hardware"], "entities": ["Gemma 4", "GizmoGuard", "ArduCam", "Raspberry Pi", "Spring Boot"], "alternates": {"html": "https://wpnews.pro/news/gizmo-guard-safeguard-bot-powered-by-gemma4", "markdown": "https://wpnews.pro/news/gizmo-guard-safeguard-bot-powered-by-gemma4.md", "text": "https://wpnews.pro/news/gizmo-guard-safeguard-bot-powered-by-gemma4.txt", "jsonld": "https://wpnews.pro/news/gizmo-guard-safeguard-bot-powered-by-gemma4.jsonld"}}