VisionAI : μια εφαρμογή Android πολλαπλών λειτουργιών υπολογιστικής όρασης εντός συσκευής, με ανίχνευση αντικειμένων σε πραγματικό χρόνο, αναγνώριση κειμένου, και γεωχωρική ανάλυση
VisionAI : a multi-function on-device computer vision Android application with real-time object detection, text recognition, and geospatial analysis
View/
Keywords
Android ; Ανίχνευση αντικειμένων ; YOLOv8 ; TensorFlow Lite ; Jetpack compose ; OCR ; ML Kit ; Firebase ; Υπολογιστική όραση ; MVVMAbstract
Smartphones have become ubiquitous sensor platforms, yet most existing applications exploit only a fraction of their perceptual capabilities. VisionAI closes this gap by offering a complete on-device computer vision suite that requires no cloud connection for its core functionality. The application embeds a quantized YOLOv8 Nano (YOLOv8n) neural network via TensorFlow Lite to perform real-time object detection at up to 30 frames per second on mid-range Android devices. Beyond object detection, VisionAI integrates Google ML Kit for optical character recognition (OCR), screen-region capture via MediaProjection, and frame-by-frame video analysis using ExoPlayer.
Detection results are stored in Google Cloud Firestore, tagged with GPS coordinates obtained from the Fused Location Provider, and rendered on an OpenStreetMap map layer. A statistics dashboard aggregates weekly activity, confidence trends, and the most frequent object categories. The application is built in Kotlin using Jetpack Compose, strictly following the MVVM architecture. It supports five natural languages (English, Greek, German, French, Spanish) and a fully dynamic dark/light theming system.
This text describes the motivation behind VisionAI, reviews the relevant literature, analyses every architectural and implementation decision, discusses the user-experience design, evaluates system performance, and presents the broad range of real-world use cases it enables — from assistive technology for visually impaired users to smart field data collection for researchers and educators.