Indoor positioning: “an image-based crowdsource machine learning approach”

Various technologies have been utilized today for recognizing client or user in the indoor areas. These technologies incorporate RSSI, Bluetooth Low Energy Beacons, Ultrasound waves, Vision-based advances, for example, fixed camera recordings QR codes, remote gadgets, etc. RSSI fingerprinting technique requires more effort and it is also expensive to be used for indoor localization frameworks working in real-time. In this research, indoor localization based on images is investigated as an option in contrast to other indoor positioning techniques using these days. Image-based indoor positioning is more affordable than RSSI based technologies being utilized. A mobile phone camera is utilized to take the pictures of area inside the building to find the user inside the building. Sensor data from various sensors isn’t required or no extra framework is required to find the client in the building utilizing indoor positioning based on an image. Microsoft Azure Custom Vision Services are utilized to locate the client; MS Azure classifies the pictures in one of the labels made. Strategy’s attainability is demonstrated by various investigations and accomplished accuracy and review is recorded above 90%. The average precision of the trained model is recorded above 95%.