Oct 2020 – AI Beyond Deep Learning – The Next Generation

  • Post author:
  • Post category:News
Wildfaces Newsletter October V8

WILDAI NEWS

October 2020

AI Beyond Deep Learning - The Next Generation

Inside this issue

AI beyond Deep Learning

Deep Learning has been made very fashionable over the last few years and people have tended to use the term synonymously with Arti￿cial Intelligence ever since “tech giants” like Google and Microsoft have embraced it.

The technology is very powerful for categorizing objects within an image. One can provide it with a large dataset of any object eg 10,000 images of dogs, and the system can be trained to detect the next dog that comes along.

Deep Learning requires large training data sets and is very computing heavy. It is di￿cult for the technology to be applied to practical use-cases. Wildfaces recently completed a project for the Hong Kong Prisons where they wanted to detect suicide attempts. There was no opportunity to collect a large dataset of relevant cases. For that requirement, Wildfaces applied its unique Reason- ingAI technology that could better mimic human intelligence, ensuring a quick and successful deployment.

– The Next Generation

Smart Tra￿c under Extreme Conditions

Second Angel Investor for Wildfaces

Comprehensive Anti-Contagion system

High

Level of Intelligence

Low

Generation 2

ReasoningAI

Generation 1

Deep Learning

 

 

 

 

 

 

 

Video tool

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

mimicking complex

 

 

 

 

 

 

 

human intelligence:

 

Image Tool

Provide Behavioral

 

Analytics &

mimicking human

Anonymous Tracking

 

Memory:

for ￿xed scenes

Categorize and segment single images

Latest vision-based AI

Generation 3

WildAI

Video tool for

Moving IoTs:

Provide ReasoningAI for moving scenes and in dynamic situations

The latest IoT camera devices are now mounted on moving carriers, for example, drones wearables or robots. Typical vision-based AI solutions can lose its context as both the foreground and the background move simultaneously. Wildfaces has overcome this challenge with a new type of AI that it has devel- oped called WildAI.

Both ReasoningAI and WildAI require very limited or even no datasets for training. This results in light computing, enabling systems to understand their environment in a way that Deep Learning never could. These are destined to ensure systems develop greater levels of comprehension than has been possible in the past.

 

OVE

R

N

P

 

 

20 YEARS

A I PR

 

 

S

 

E

IC

 

ACT

 

 

Computing

No dataset

Quick

Practical

Light

for training

Deployment

Approach

Smart Tra￿c under Extreme Conditions

Wildfaces has just completed the ￿rst phase of a Smart Tra￿c Management System in Hong Kong. The requirement was theoretically simple to detect trespassing on the freeway. However, the weather conditions were extreme and the lighting conditions were poor and inconsistent resulting in many false alarms for systems that had been tried out before.

These previous systems generated 200 to 1,000 false alarms per day per camera which was unacceptable for the users who were running a mission critical system. While the accuracy of the system was important the ability to eliminate false alarms was also essential.

The Wildfaces system achieved 100% accuracy during the ￿rst 8 weeks of operation with only one false alarm during that period, thanks to its unique Arti￿cial Intelligence based NAMS (Nuisance Alarm Minimization System).

The most important factor was ease of deployment. With Wildfaces’ ReasoningAI no massive datasets were required for training, no GPUs were needed in the network and it was possible to deploy the system in days rather than months.

This HTML is created from PDF at https://www.pdfonline.com/convert-pdf-to-html/

Second Angel Investor for Wildfaces

A second Angel has now invested in Wildfaces.

“A unique value proposition, a new perspective and a pragmatic approach for vision-based analytics. WildAI has the potential to be one of the leading foundation platforms.”

The family o￿ce of Mr. Ken Or has now invested in Wildfaces. He believes in WildAI, not only because of its advancement from Deep Learning but because it can o￿er a wider range of appli- cable use-cases alongside the emerging development of IoT devices.

To guide it on its path, Ken has personally joined the company as its Director of Strategy, respon- sible for corporate development and capital raising. Beyond looking for external monetary fund- ing, he will steer the company to prioritize business development synergies with investors. The company may potentially o￿er project-based capital raising rounds for its research & develop- ment initiatives.

Mr. Ken Or

Meet Mr. Ken OR and our Management at this Virtual CVCF event.

3-4 Nov

Email us at bd@wildfaces.ai for meeting arrangements

Comprehensive Anti-Contagion system

As organizations attempt to go back to normal after the peaks of the pandemic they need to ensure that their sta￿ and visitors remain safe. To help them Wildfaces has used its unique ReasoningAI technology to develop a comprehensive suite of tools which include:

An automated Fever Detection capability which operates without the need to have paramedics watching every camera.

A contactless Access Management System that provides multi-factor authentication and access for busy environments without people having to press buttons or touch anything.

An evacuation management system that operates at close to 100% accuracy.

A Social Distancing system that will raise alarms if social distancing norms and not met.

A PPE detection system which will advise if people are not wearing their masks at all times.

All these systems are computing light and require no training data sets.

Dwelltime

Evacuation

Access

Contactless

PPE

SocialDistancing

Fevercheck°

For more information give us a visit! www.wildfaces.ai | bd@Wildfaces.ai © All Rights Reserved. WildFaces Technology Ltd

This HTML is created from PDF at https://www.pdfonline.com/convert-pdf-to-html/