BrainChip September 2018 Quarter Update 1 Brainchip OCTOBER 2017 |
Disclaimer This presentation is not a prospectus nor an offer for securities in any jurisdiction nor a securities recommendation. The information in this presentation is an overview and does not contain all information necessary for investment decisions. In making investment decisions in connection with any acquisition of securities, investors should rely on their own examination of the assets and consult their own legal, business and/or financial advisers. The information contained in this presentation has been prepared in good faith by BrainChip Holdings Ltd, however no representation or warranty expressed or implied is made as to the accuracy, correctness, completeness or adequacy of any statements, estimates, opinions or other information contained in this presentation. To the maximum extent permitted by law, BrainChip Holdings Ltd, its directors, officers, employees and agents disclaim liability for any loss or damage which may be suffered by any person through the use or reliance on anything contained in or omitted in this presentation. Certain information in this presentation refers to the intentions of BrainChip Holdings Ltd, but these are not intended to be forecasts, forward looking statements or statements about future matters for the purposes of the corporations act or any other applicable law. The occurrence of events in the future are subject to risks, uncertainties and other factors that may cause BrainChip’s actual results, performance or achievements to differ from those referred to in this presentation. Accordingly, BrainChip Holdings Ltd, its directors, officers, employees and agents do not give any assurance or guarantee that the occurrence of the events referred to in the presentation will actually occur as contemplated. 2 Brainchip OCTOBER 2017 |
September 2018 Quarter Update Announcement of Akida TM Neuromorphic System-on-Chip (NSOC) Architecture Partnership with Telesikring to deliver AI-powered Video Analytics to Police and Security Customers across Scandinavia Acquistion of License to Cybersecurity technology from Democritus University of Thrace, Greece Gaming Partners International (GPI) Demonstration of Integrated Vision System at the G2E gamining conference in Las Vegas, Nevada 3 Brainchip OCTOBER 2017 |
September 2018 Quarter Cash Receipts The Company finished the September 2018 quarter with a cash balance of US$9.995M. Cash receipts for the quarter were US$246,000. Headcount at the end of the quarter was 35 fulltime employees and two fulltime contractors in sales and marketing. In the June quarter the Company invoiced SNTech US$609,135 based on an existing Licencing and Development Agreement. The Company has initiated a full audit of SNTech’s books, records and source code. The results of the audit are expected in November 2018. 4 Brainchip OCTOBER 2017 |
December 2018 Quarter Outlook The Company expects an increase in cash receipts from OEMs, System Integrators and other partners in the December quarter of 2018. The Company’s sales pipeline continues to grow and now includes over 500 leads, 105 active accounts, 55 qualified opportunities and 17 design wins. The Company currently supports approximately 21 committed or active trials for which most are covered by Non-Disclosure Agreements. 5 Brainchip OCTOBER 2017 |
BrainChip Application Update Automotive Automobile Manufactures Third Party Vendors Civil Surveillance Original Equipment Manufacturers End-Users and System Integrators Gaming Neuromorphic computing for multiple markets Vision systems Cybersecurity Financial Technolog Agricultural Technology 6 Brainchip OCTOBER 2017 |
Akida Market Penetration Company will provide investors insight with regard to activity in the following markets: Industrial – Internet-of-Things (IoT), Agricultural Technology, Financial Technology, Cybersecurity, Robotics, Process Control and Manufacturing Automation, , Surveillance, Gaming Automotive – ADAS and AV Consumer – Internet-of-Things (IoT), Cell Phones, Personal Computing and Home Security 7 Brainchip OCTOBER 2017 |
BrainChip and the AI Timeline 8 Brainchip OCTOBER 2017 |
Semiconductor Industry Disruption AI Acceleration Chipset Forecast 70,000 60,000 Training Inference General Purpose 50,000 40,000 $M 30,000 20,000 10,000 0 2018 2019 2020 2021 2022 2023 2024 2025 9 Brainchip OCTOBER 2017 | Source: Tractica Deep Learning Chipsets, Q2 2018
Neural Network Comparison Convolutional Neural Networks Spiking Neural Networks Characteristic Result Characteristic Result Computational Matrix Multiplication, Math intensive, high Threshold logic, Math-light, low functions ReLU, Pooling, FC power, custom connection power, standard layers acceleration blocks reinforcement logic Training Backpropagation off- Requires large pre- Feed-Forward, on or Short training chip labeled datasets, long off-chip cycles, continuous and expensive training learning periods Math intensive cloud compute Low power edge deployments 10 Brainchip OCTOBER 2017 |
Studio 11 Brainchip OCTOBER 2017 |
BrainChip Studio Development History Custom software deployments Gaming Civil surveillance Machine vision Mohegan Sun French Homeland Security (DGSI) SAFRAN augmented reality 10 years of development Las Vegas Trials French National Police (Paris, Toulouse) Real-time card and Forensics and live analytics Object recognition chip detection Mass-Market Product 12 Brainchip OCTOBER 2017 |
Challenges Facing Forensic Video Analysis Exabytes of previously recorded video Low resolution and noisy environments Prohibitively expensive to “Classify” all video using deep learning 13 Brainchip OCTOBER 2017 |
BrainChip Studio Features and Benefits Highly effective object Fo Forensic Object One-Shot learning to recognition rapidly search for Se Sear arch technology in patterns/suspects low-resolution environments Fa Face Detection Automatic face Facial classification of and and detection and top matches extraction Classificati Cl tion Targeting the civil defense/police marketplace Force multiplier: Single Uses existing low-fidelity Find suspects across hundreds of individual can do the scanning of infrastructure for live or videos many recorded forensic search 14 Brainchip OCTOBER 2017 |
Technology – BCS Learn BCS Learn It’s all about shape recognition, not biometrics and key points! RECOGNIZE BrainChip LEARN Recognition of patterns Visual target recognition with similar visual to identify model features 15 Brainchip OCTOBER 2017 |
BrainChip Accelerator Benefits BrainChip Accelerator FPS Increase Compared to Software Only 6x 6x 5x Improved probability of object detection Fast video scaling Faster time to results More channels, faster 3 Scales, 1 Model 3 Scales, 4 Models 6 Scales, 4 Models Lower total cost of ownership Lower initial cost, reduced power, and longer MTBF 16 Brainchip OCTOBER 2017 |
17 Brainchip OCTOBER 2017 |
BrainChip Akida TM NSoC World’s first Neuromorphic System on Chip (NSoC) Efficient neuron model Innovative training methodologies Everything required for embedded/edge applications On-chip processor Data->spike conversion Scalable for Server/Cloud 18 Brainchip OCTOBER 2017 |
Akida NSoC Architecture 19 Brainchip OCTOBER 2017 |
Akida Neuron Fabric Most efficient spiking neural Right-Sized for embedded network implementation applications 1.2M Neurons 10B Synapses 10 classifiers (CIFAR 10) 11 Layers Able to replicate most CNN functionality 517K Neurons Convolution 616M Synapses Pooling Fully connected Meets demanding performance criteria 1,100 fps CIFAR-10 82% accuracy 20 Brainchip OCTOBER 2017 |
Efficient Neuromorphic Computing Fabric Keys to efficiency Relative Implementation Efficiency (Neurons and Synapses) Fixed neuron model Right-sized Synapses minimized 300X on-chip RAM 6MB compared to 30-50MB Programmable training and firing thresholds Flexible neural processor cores Highly optimized to perform convolutions Also fully connected, pooling Efficient connectivity Global spike bus connects all neural processors 3X Multi-chip expandable to 1.2 Billion neurons 21 Brainchip OCTOBER 2017 |
Neuromorphic Computing Benefits ~$1,000 Cifar-10 IBM 83% Tremendous TrueNorth Cifar-10 82% ~$10 throughput with low BrainChip Top-1 Accuracy Akida power 6K fps/w 1.4K fps/w Math-lite, no MACs No DRAM access Cifar-10 ~$1,000 80% for weights Cifar-10 Xilinx 79% ~$10 Intel ZC709 Myriad 2 Comparable accuracy 6K fps/w 18 fps/w Optimized synapses and neurons ensures precision Frames per Second/watt 22 Brainchip OCTOBER 2017 |
Akida NSoC Applications 23 Brainchip OCTOBER 2017 |
Akida Development Environment 24 Brainchip OCTOBER 2017 |
Akida Introduction World’s first NSoC Low power and footprint of neuromorphic computing Highest performance /w/$ Complete solution for embedded/edge applications – but scalable for cloud/server usage Sampling 2H 2019 25 Brainchip OCTOBER 2017 |
Recommend
More recommend
