PRELIMINARY DESIGN REVIEW Team 1 October 19, 2018
MirrAR Taehwan Chon Jyerrmee Nerona John Nguyen Kevin Pierre
Advertising Solutions Today Current marketing industry stagnant Main efforts towards Social Media and Internet Retail e-commerce sales worldwide forecast to nearly double between 2016 and 2020.
Current Advertising Industry In-store experience still relevant 71% prefer in-store for jewelry 68% prefer in-store for apparel Consumers prefer convenience Low incentive to go to physical storefronts
Context: Effect on Consumers When buying glasses, many times items may not be in stock for different Colors Engravings Models Sizes
Requirements Analysis: Specifications Real-Time augmented reflections of user Wide angle camera - 1080P Resolution Effective range 5 feet 24 inch TN monitor Refresh rate 60Hz Facial recognition for preferences and history Fast processing to produce smooth movements and clear visuals ~10ms latency
Requirements Analysis: Inputs and Outputs Camera Edge Detection Data Eyewear model Output Live video playback (1080p) Eyewear reflected on user’s face and adjusted based on angle and size of the face
Design Alternatives Sephora Augmented Reality Mirror Amazon AR Mirror ModiFace’s 3D Mirror All angle display Amazon AR Mirror Camera to track user’s body Mirror body consists of a reflective panel and a screen behind it that projects the image
Design Alternatives Sensemi’s Sense Magic Mirror 360 degree base movement to for height adjustment Maya Software for 3D Models External USB for updates Indestry’s Broadcast AR Minimum distance = 6 meters (20 feet) Screen consists of three layered elements: LED Screen, Plasma Wall, and Projection System
Block Diagram
Our Solution: Hardware Camera Live video streaming High Frame Rate & Clarity High Data Throughput Ex: Logitech C922 1080p, 60fps Display High Clarity & Low Delay 24 inch Android Smartphone Application Easy to use UI Communicates via Bluetooth
Our Solution: Hardware Cont’d. Custom PCB Two different processors Running program tasks in parallel Maximizing throughput, minimizing processing time, minimizing delay Bluetooth Component Communicate with smartphone
Sensors Camera Compromise between high clarity, high data throughput, and reasonable price Images run through facial recognition algorithm
Block Diagram
Our Solution: Software Open Source Computer Vision Library (OpenCV) Library aimed at real-time computer vision Motion-Detection Augmented Reality Facial Recognition Object detection using Haar Cascades Pre-trained classifiers for facial features Languages Python and C++
Our Solution: Software Cont’d. Head Pose Estimation for Angled Vision TensorFlow and OpenCV TensorFlow: Library for dataflow programming and machine learning Three steps 1) Face Detection 2) Facial Landmark Detection 3) Pose Estimation
Processor Specifics and Requirements Goal is to be able to project glasses on face, will need enough power for this Real life example = Snapchat Requirements for Snapchat on mobile: CPU – 0.2 Processing Power GPU – 1.4 Graphical RAM – 1.2 OS – Android 2.2/ iOS 5
Power Requirements Most high end desktops with monitors use around 100-300 Watts. Given this we will want to use a power supply that can provide >=300W: The power supply that fits this criteria is a 100-240VAC Power Supply.
UI Android application using Android Studio Language: Java Inventory database accessible through application Every interaction with mirror executed through the application
Communication Between App and Board App will communicate with board via Bluetooth Phone will have built in Bluetooth Once app is created, phone will be able to connect HC-05 Wireless Bluetooth Slave and Master Module
MDR Deliverables Demonstration of facial recognition Screen will reflect user and box around their face Demonstration of application communication Android application displays image of glasses on MirrAR
Thank You Questions