bigjames

SceneGuard AI is an AI-powered workplace safety inspection platform that uses multimodal artificial intelligence and computer vision to identify potential hazards in warehouses and industrial environments. By combining Meta's Muse Spark 1.3 and SAM 3.1 models, SceneGuard transforms ordinary images and live camera feeds into actionable, evidence-based safety assessments. The Problem:- Warehouses and industrial workplaces face ongoing safety risks from blocked walkways, obstructed fire lanes, and improperly stored materials. These hazards can contribute to slips, falls, and forklift-related accidents. Traditional inspections rely heavily on manual observation, making them time-consuming, intermittent, and susceptible to overlooked hazards. Our Solution:- SceneGuard AI enables users to upload workplace photographs or use a live camera to detect potential hazards and receive annotated visual findings, severity assessments, and recommended corrective actions. Our AI inspection pipeline works in four stages: 1. Scene Understanding: Meta's Muse Spark 1.3 analyzes the image, identifies potential hazards, and interprets environmental context, including safety signage and designated walkways. 2. Object Localization: Meta's SAM 3.1 uses text-guided prompts to locate relevant objects and generate precise bounding boxes. 3. AI Verification: Muse Spark evaluates the detected objects against the original scene to determine whether they actually present an obstruction. For example, boxes properly stored on warehouse shelves are distinguished from boxes blocking an aisle. 4. Risk Assessment: A rule-based engine assigns severity levels, corrective recommendations, and human-review flags to verified findings. Tech Stack: React, TypeScript, Tailwind CSS, Python, FastAPI, and Meta Model API (Muse Spark 1.3 and SAM 3.1). Our Goal: Make workplace safety inspections faster, more accessible, and proactive while supporting human safety professionals rather than replacing them.