I trust that most shareholders are aware that on 14 December,2019 Brainchip Engineers jointly presented the AKIDA technology platform withEngineers from TATA Consulting Services (TCS) doing real time gesturerecognition.Most will also remember that supporting the demonstration were a number of visual aids/slides prepared by now Dr. Arijit Mukherjee at TCS whereon he stated that he was looking forward to building a robotic platform based on the AKIDA chip.If you use the below link it will take you to the Robotics group site at TCS and there you will find a list of all current members and then a running list of all papers prepared and published by that group since 2016. It is a very long list that does not run in perfect date order but I have extracted the following papers which confirm that from 2019 into 2020 Dr. Arijit Mukerjee continued to work on this use of neuromorphic computing.
Following this list I have also included a link to TATA Elxsiwhich seems to take over this work and you should open the link provided as itis a much easier format to follow however that which it will allow me to copyhas been reproduced and I suggest that as you work your way through the listyou will be taken by how what they have moved forward to work on perfectlyaligns with all of the recent presentations and videos put out by Brainchip.
Is TATA India’s largest company still working with Brainchip andis it one of its EAP customers I will let you decide???
My opinion only DYOR
FF.
AKIDA Ballista (to Bollywood and back.)
TATA CONSULTING SERVICES – BuildingNext Generation Smart Machines
https://www.tcs.com/building-next-generation-smart-machines
*************************************************************************************
Human Gesture Recognition using Spiking Input on Akida NeuromorphicPlatform
Neural Information Processing Systems
12/10/2019
Sounak Dey, Arijit Mukherjee,
****************************************************************************
Application of Spiking Neural Networks for Action Recognitionfrom Radar Data
The International Joint Conference on Neural Networks
21/03/2020
DIGHANCHAL BANERJEE, Sounak Dey,Smriti Rani, Tapas Chakravarty,Arijit Mukherjee, Arijit Chowdhury, ARUN M GEORGE,
****************************************************************************
A Reservoir-based Convolutional Spiking Neural Network forGesture Recognition from DVS Input
The International Joint Conference on Neural Networks
21/03/2020
Sounak Dey, ARUN M GEORGE, DIGHANCHAL BANERJEE, ArijitMukherjee,
**************************************************************************
Towards Spike based Human Gesture Recognition on AkidaNeuromorphic Platform
International Conference on Distributed Computing and Networking
24/12/2019
Sounak Dey, DIGHANCHAL BANERJEE, Arijit Mukherjee, ARUN M GEORGE,
https://tataelxsi.com/storage/insights/December2020/c2IBlwKiKdz2KfBnSJOj.pdf
Scope
• Porting real time AI applications on
MCU Platform
• Optimizing AI algorithms for porting
on low memory footprint devices
• AI based Gesture detection, User
Posture recognition, Sound
Analytics, Expression detection,
Footstep detection etc.
• Optimization and simplification of AI
inference engines retaining the
accuracy
• Creating a demo setup to showcase
the capabilities
AI ENGINE ON EDGE PLATFORM – SMART HOME APPLICATION
Enabling intelligence on an edge Microprocessor to assist cloudbased AI engine
Challenges
• Real time response
• Data set collection, cleansing, and
segregation
• Optimization of algorithm to run
on an embedded system
Tools & Framework
• TensorFlow, TensorRT
• Keras
Scope
•Porting Ultra light DNN algorithm on a microcontroller
•Analysis of Conveyer belt operation, robotic gear movement
•Using rotary vibration history data for predicting abnormal and normal gearoperations
•Implementing Predict Failure and triggering Schedule maintenance
EDGEAI FOR PREDICTIVE ANALYSIS - INDUSTRIAL AUTOMATION
EnablingAnalytics at the Edge for Wafer Production Process Challenges
•Motor vibration analysis under different load conditions
•Anomaly detection in less than 0.10 sec Tools & Framework
•TensorFlow, TensorRT
•Keras, PyTorch
(20min delay)