THE EDGE AI FABRIC COMPANY J U L Y 1 6 , 2 0 1 8 Poi ntR Data I nc. 181 2 nd St. San Franci sco, CA 94105
POINTR EDGE AI FABRIC FOR APPLIED AI AI is a competitive differentiator Edge AI fabric integrates sensors and cameras in a mesh network providing for enterprise to drive productivity gains in core businesses real-time business intelligence at scale • Moves inference to end points transmitting The main catalyst for AI adoption only meta-data to the cloud is harvesting unstructured data sources to provided real time Minimizes cables, bandwidth choke business insights points, and expensive compute power Velocity and volume of the data consumption making it affordable and produced is too large to process scalable in the cloud The software virtualizes hardware PointR EDGE AI FABRIC variability and provides self-provisioning Connects cameras and sensors and healing into an inference fabric 2 0 1 8 T h i s M a t e r i a l C o n t a i n s P o i n t R D a t a C o n f i d e n t i a l I n f o r m a t i o n 0 7 . 0 3 . 2 0 1 8 2
RETAIL USE CASE In-store Automation – Amazon GO Example Shoppers “grab-and-go” removes cashier choke points Integration with e-commerce Shelf inventory, price check and planogram compliance Shrinkage Surveillance Shopper Marketing 2 0 1 8 T h i s M a t e r i a l C o n t a i n s P o i n t R D a t a C o n f i d e n t i a l I n f o r m a t i o n 3
THE APPLIANCE 4
THE POINTR DISRUPTION Specs Equals 21x Xeon 2790v4 CPU 12 AI Accelerators Dell r730 servers Cluster Processing 120 Watts Peak 1/100 th power 24 Teraflops @16-bit PCIe Fabric 1/20 th price @10Gbps DMA ops That’s 21 Dell servers! 2 0 1 8 T h i s M a t e r i a l C o n t a i n s P o i n t R D a t a C o n f i d e n t i a l I n f o r m a t i o n 5
HOW DOES IT WORK? In the Field Data Center Data Collection: millions of images for thousands of labels 1-click publish Cameras 2 1 Step 2 USE the Models trained by PowerAI Step 1 AC922 NEWELL TEACH by feeding data Instant Real Time Response PointR Field Appliance Step 4 3 4 Product Updates from stores require re-training Step 3 Daily Weekly Updates Information changes fed Dashboard alerts manager of what the camera sees back to data center 6
ENTERPRISE VERTICALS Retail Transportation Manufacturing Healthcare Surveillance GrabandGo Railroad Maintenance Inspection Medical Imaging City Traffic Monitoring Loss Prevention Fleet Safety with in-cab Quality Control Pathology and Threat Matrix behavior Radiology Out-of-Stock Precision Alignment 2 0 1 8 T h i s M a t e r i a l C o n t a i n s P o i n t R D a t a C o n f i d e n t i a l I n f o r m a t i o n 7
POWERAI VISION INTEGRATION WITH POINTR Live Video PowerAI Vision PointR Data Label GUI New AI Model Train AI Models New Video Deploy Event Detection Training Data AC 922 Newell Edge Inference Power Server 2 0 1 8 T h i s M a t e r i a l C o n t a i n s P o i n t R D a t a C o n f i d e n t i a l I n f o r m a t i o n 8
POINTR POWERAI BUNDLES 2 0 1 8 T h i s M a t e r i a l C o n t a i n s P o i n t R D a t a C o n f i d e n t i a l I n f o r m a t i o n 0 7 . 0 3 . 2 0 1 8 9
RETAIL USE CASE In-store Automation – Amazon GO Example Shoppers “grab-and-go” removes cashier choke points Integration with e-commerce Shelf inventory, price check and planogram compliance Shrinkage Surveillance Shopper Marketing 2 0 1 8 T h i s M a t e r i a l C o n t a i n s P o i n t R D a t a C o n f i d e n t i a l I n f o r m a t i o n 10
TRANSPORTATION USE CASE Multi Billion Dollar Market (Railroads) Inspect Pantographs for French railroad Maintenance 60 annual failures in Paris cost millions in euro’s and incalculable loss in public confidence affecting SNCF political reputation I nspect tracks for BNSF and Union Pacific Yard surveillance, connected trains, cargo liability inspections Autonomous Railway 2 0 1 8 T h i s M a t e r i a l C o n t a i n s P o i n t R D a t a C o n f i d e n t i a l I n f o r m a t i o n 0 7 . 0 3 . 2 0 1 8 11
MANUFACTURING USE CASE Predictive Maintenance Quality Control Precision Alignment 3D Vision Inspection Just one minute of Inspect products coming Poorly trained workers Market-and-Market downtime in an off the assembly line can be augmented report suggests that automotive factory can for compliance and with machine the overall machine cost as much quality guidance vision market could be as $20,000 on high- worth up to $14.43 profit vehicles Check for cap closures, billion by 2022 for defects, position, and manufacturing Perform real-time AI labels automobiles from cameras attached to robots 12 2 0 1 8 T h i s M a t e r i a l C o n t a i n s P o i n t R D a t a C o n f i d e n t i a l I n f o r m a t i o n
CALL TO ACTION The Problem The Solution Peak awareness but few implementations Pre-baked Appliance for Deep Learning Scarce data science talent ($1M salary) Rapid onboard customer use case Today dev that takes 9 months and millions of dollars Ongoing consulting resources are very expensive ($10M+) we shrink to 3-6 weeks Proof-of-Concept Fixed-time/Fixed Bid for one customer use case Expand use cases and footprint with dedicated team 2 0 1 8 T h i s M a t e r i a l C o n t a i n s P o i n t R D a t a C o n f i d e n t i a l I n f o r m a t i o n 13
THANK YOU 2 0 1 8 T h i s M a t e r i a l C o n t a i n s P o i n t R D a t a C o n f i d e n t i a l I n f o r m a t i o n 14
Recommend
More recommend
