📱 Real-time room scanning & object change detection using YOLOv8 — optimized for mobile and edge devices.
WhatsApp.Video.2025-04-06.at.10.11.22_b45884ce.mp4
ScanFlow is an AI-powered mobile application that helps track and detect spatial changes in dynamic environments such as hotel rooms, warehouses, retail shelves, and office spaces. It scans a room, builds a baseline map, and detects changes like missing, moved, or newly added objects in subsequent scans.
🎯 Designed for mobile, it empowers users with efficient, lightweight, and actionable spatial insights for inventory, security, and space optimization.
- Initial Room Mapping – Create a baseline by scanning the room and storing object layout.
- Change Detection – Detect & categorize new, removed, or repositioned items using YOLOv8.
- Optimized Edge Processing – Processes only changed areas to save resources.
- Mobile/Edge Deployment – Runs efficiently on low-power edge devices or phones.
- User-Friendly Visualization – Heatmaps, highlights & difference view overlays.
- Object Categorization & Alerts – Triggers alerts when specific objects go missing or move.
- Multi-Scan Analysis – Tracks long-term changes and trends across multiple visits.
- External System Integration – Connects with inventory/security/facility software.
- 🗣️ Calling Agent (Smart Dirt & Damage Advisor)
An AI-powered assistant that analyzes cleanliness/damage and calls the manager (voice/text) to report critical issues.
Example alerts:"Room 104 has visible carpet stains and clutter. Immediate cleaning required."
"Room 212 is clean. ✅ No issues detected."
- Streamlined check-in/checkout with photo comparison.
- AI detection for missing/damaged/moved items.
- Auto-generated issue reports.
- View room history & logs.
- Review and verify AI alerts.
- Dashboard for trends and analytics.
- Confirm or override AI insights with context.
- Strategic decision-making support.
- Receives automated calls/texts from the Calling Agent for high-priority rooms.
- ⚡ Real-Time Inventory Management
Instant tracking of inventory movement and updates. - 🎯 Advanced Object Detection
Fine-tuned YOLOv8n model for indoor/hotel-specific objects. - 🔐 Person Identification (Optional)
Presence detection for security monitoring. - 🧩 Customizable Architecture
Extendable to any space – from rooms to entire buildings. - 🗣️ Calling Agent for Actionable Alerts
Voice/text AI assistant for instant room prioritization.
| Component | Technology Used |
|---|---|
| Object Detection | YOLOv8n (fine-tuned) |
| Backend | Python, Flask/FastAPI |
| Model Training | Ultralytics, Roboflow |
| Frontend/App | React Native / Flutter |
| Deployment | Android, Raspberry Pi |
| Visualization | OpenCV, Matplotlib, Gradio |
| Audio Alerts | gTTS, pyttsx3, ElevenLabs |
| Dataset | Custom Hotel/Indoor Dataset |
- Model: YOLOv8n (nano variant)
- mAP@0.5: 55–65% (small objects)
- mAP@0.5:0.95: 29.97%
🔍 mAP@0.5:0.95 measures detection consistency across IoU thresholds. Scores are competitive for lightweight models and will improve with further tuning.