2011 - CTO, Presenting Suplea

Suplea was a machine vision Startup where state of the art algorithms were tried to be monetized.

Transcript

1. S U P L E A S U P L E A Real­time pattern recognition and learning jernej@suplea.com vlado@suplea.com
2. Vast amounts of unclassified objects
3. lamp sofa sofa table chair window people globe bus street lamp microwave chair pipe cupboard plate painting teapot stand table table table laptop chair chair chair books bus poster reflectors bird s building banner street lamp people bookcase cabinet washing machine Indexing by content with Suplea tech
4. Jeff Hawkings, co-founder Palm and Handspring: - created the Redwood Neuroscience Institute (RNI) - developed a theory of neocortex (J.Hawkings, On Intelligence, 2004) - founded Numenta in 2005 to develop a technology platform derived from his theory Comparable technologies
5. Our approach Learning Prediction Hierarchical and unsupervised learning In real time Incremental learning of multi-class object representations With proven experimental results
6. Only 1.6Mb to store a 15- class vocabulary on disk Training time for each additional class is reduced while learning multiple classes incrementally We can detect: multiple objects in real time And our technology is scalable: Suplea Vision
7. TAM SAM SOM Est. market growth Future predictions Needed features Data mining, BI, financial market, weather forecast $10B $7B $5B 20% $14B in 2014 for BI only 4.1 Algorithm generalization Online video advertising market $2.3B in 2010 2.1 $2B $1B 43% 2.1 in US only (2010) $11.3B in 2012 2.1 Color dimension, Differentiating subcategories Smart cameras / application-specific machine vision (ASMV) systems $2.9B in 2010 1.1 $2B $500M 10.4% 1.1 $4.2B in 2015 1.1 Time dimension, autonomous learning of new categories Ipad application and mobile interfaces opportunities $7B in 2009 $100M $50M 30% $17B in 2012 3.1 Color dimension 1.1: http://www.bccresearch.com/pressroom/IAS010C.html 1.2 http://www.automationworld.com/webonly-2071 1.2 Machine Vision Markets—2005 Results and Forecasts to 2010 2.1: http://techcrunch.com/2008/08/13/lehmans-online-us-advertising-forecast-another-20-billion-in-growth-by-2012-and-onlin e-video-takes-off/ 3.1: http://mashable.com/2010/03/17/mobile-app-market-17-5-billion/ 4.1 http://www.forrester.com/rb/Research/state_of_business_intelligence_software_and_emerging/q/id/56749/t/2 Market opportunities
8. Technology Roadmap Color dimension Optimized object classification Differentiating subcategories Time dimension Subcategory classification Self learning of new shapes Data mining Analyzing time sequences Algorithm generalization Autonomous knowledge gathering 2011 Q1 2011 Q2 2011 Q3 2012 Q1 2011 Q4 Technology development Technology placement
9. Predictable space/time execution, appropriate for real time problems Complex context mining via multiple object detection Viable approach towards high degree of robot autonomy Cherry picking apps: Image/video service that searches by content and displays relevant ads Visual search and touch aided design Smart cameras Applications
10. Beyond Smart Cameras Systems build on Smart Cameras grids SoC design for application- specific machine vision High level of flexibility using different object libraries No additional CPU required Real time execution
11. NextGen touchscreen interface Express complex finger gestures in visual search Bring intuitive UI to the next level User gestures as actions Responsiveness and alertness
12. Delivering relevant ads 80% of Internet video streams are currently not monetized Pictures without relevant tags or context are no good Suggesting and Searching by content
13. The Team Ales Leonardis, Prof., PHD. founder is a full professor and the head of the Visual Cognitive Systems Laboratory with the Faculty of Computer and Information Science, University of Ljubljana . Marko Boben, PHD. founder Is an assistant for mathematics at Faculty of Mathematics and Physics, University of Ljubljana, Slovenia. Sanja Fidler, PHD. founder is a doctoral student with the Faculty of Mathematics and Physics, University of Ljubljana. Kristjan Pecanac. Three times founder, intern CEO Vlado Dimitrieski, founder, applications development, graduated in computer science at University of Ljubljana, Slovenia Jernej Makovsek, founder, technology development undergraduate at University of Ljubljana, Slovenia studying physics. Žiga Ham, web development, currently an undergraduate at University of Ljubljana, Slovenia, studying computer and information science. University staff Developers
14. For startups and businesses who are using pattern recognition techniques We provide a technology platform that works across multiple domains Unlike other technology providers who have scalability issues and/or are locked on a single domain Our Company offers a single platform that works across multiple domains. The advantage Users are getting results in real-time and don't have to invest in big hardware or research to realize their goals. Summary