SenSen Networks October 2018
• Introductions • Background of SenSen Agenda • SenSen's Video Analytics, AI and Machine Learning Capabilities • SenGAME solution for Table Analytics • Integration Options
SenSen Networks Overview Melbourne Australia HQ based company. Spun out of University of Who technology sydney Listed on Australian Stock Exchange (ASX: SNS). Market Cap $60M we We have ~60 people spread between Melbourne, Sydney, are Singapore, India and Canada We specialize in AI enhanced process automation software solutions World leading expertise in Computer vision, deep machine learning, data analytics, Artificial intelligence
SenGAME Solution AI enhanced Video Analytics for Table Games
Problem • For every game played on a gaming table determine • Game Begin Time • Game End Time • Time Between Games • Number of Active Players (Occupancy) • Type of Bets placed • Main bet • Back bet • Side bets • Value of each bet • Track cards to determine game outcome • Track Cash exchanges • Track Float in real time • Bet Attribution to rated and unrated players
The Solution • High resolution cameras from Android Phones • Video/Image Analytics using GPU based accelerated processing • Deep Machine learning software to learn casino specific layouts, chip colors & patterns. • Real-time data available in centrally hosted SQL data base • REST API and SQL for easy integration with data warehouses and business intelligence & analytics applications Video Processing Unit
Game Begin & End • Uses cards appearance and disappearance to determine Game Begin and Game End • Accuracy ~ 100% • Gracefully handles exceptions
Number & Type of bets • Detect image coordinates of bets placed on the gaming table. • Check which player area the bet falls in to determine if a particular player is active • Check for Perfect pair bets, main bets, back bets, side bets and etc., using region specific polygons within the camera field of view • Determine the total number of active players areas (also referred to as table occupancy) • Accuracy ~ 100%
Bet value • Patent pending technology • Works with single and multiple chip stack including dirty stacks that contain multiple colored chips within the same stack. • Accuracy ~ close to 100% • Accuracy may drop in crowded occlusion prone scenarios • Accuracy is limited only by operationally rare, perfect occlusions from other chip stacks or objects
Product Road Map • Track cards to determine game outcome • Track Cash exchanges • Track Float in real time • Bet Attribution to rated and unrated players
Key Features Retrofit to Easy to set High Rapid Flexible IP Low cost existing up & Accuracy training architecture tables Configure Processing can be done on the table using a mini pc or centrally with one powerful server Accuracy is No need to change Protected by improved Typically takes 2-3 layouts, or chips or Supports varied Android phones Typically takes less multiple global continuously by weeks to learn new tables and no need mounting options and off the shelf than 1hr to set up patent (pending) deep casino chips and to provide and positions to mini PC a table applications reinforcement layouts additional dealer optimise accuracy learning training Future roadmap to process everything on the phone
Crown Casino Installation
Applications • Hands per hour • Dealer KPI measurements • Activity/occupancy heat maps • Accurately rate players • Loyalty • Bonus bets or Jackpots • Floor Yield Management • Fraud Detection • Cash drop
• REST API Integration • SQL Options • JSON • XML over TCP/IP
Product Development Status & Road Map • SenGAME is developed in partnership with Crown Casino in Melbourne • 3 years of development • SenGAME 1.0 used surveillance cameras • SenGAME 2.0 used Microsoft Kinect & 3D Time of Flight Cameras • SenGAME 3.0 (Current version) uses Android Phones retrofitted to tables • Current features • Game Begin & End (Accuracy ~ 100%) • Bet count and Bet Type (Accuracy ~ 100%) • Number of players or Occupancy (Accuracy ~ 100%) • Bet Value (Clean & Mixed Stack) (Accuracy ~ 98%-100%)
SenGAME – Steps to make it work in a new casino Collect Data Install from Rapid Deploy & Set up & Retrain Cameras Gaming training Test Configure Floor Enjoy the Magic Feedback the Set up each table Mount the Record Videos and Deploy the errors into the Train a deep neural with appropriate cameras as per images when software and test it networks to network game logic and run recommendations games are played in live environment improve the results live if necessary
Recommend
More recommend
