The Neurala Difference

Massimiliano Versace,

Neurala Founder and CEO in 2014 TEDx Talk.


From the beginning, Neurala pioneered deep learning neural network software that could be deployed on ordinary processors with low-cost sensors. We like to say that our approach enables our AI to run anywhere. We can give great results on toys and outstanding performance on high-end GPUs.

The difference is in our DNA. Our first project for NASA was to design AI for autonomous planetary exploration. Super computers were not available. Battery life was limited. Fast internet access was impossible. Our deep learning neural networks had to be lightweight and perform in real time without ground intervention.

To do that, we created The Neurala Brain, highly efficient software which is based on the way brains work in nature. Today’s commercially-available Neurala Brain builds upon that effort. We achieve industry-leading performance on systems that range from single-board computers to cloud-based servers.

DEEP LEARNING

The Neurala Brain learns quickly from examples, not pre-programmed rules.


NEURAL GPU PROCESSING

Handles multiple sensors using low-cost graphics processors (U.S. Patent No. 8,648,867).


Neurala’s Technology Makes Machines More Useful

FAST LEARNING

Learns quickly and adapts in real-time.

HIGH PERFORMANCE

3x faster than competitors on an industry test

SCALABLE

Support ranges from SBCs to powerful GPU servers.

EASY-TO-USE

Use just few lines a code to run on standard hardware.

FEATURE

Real-time “what and where”

DESCRIPTION

A product with The Neurala Brain and the Brains for Bots SDK can find and recognize objects and concepts in real-time. 

BENEFIT

Drones can act while in flight. Cars can know what is nearby. Security systems can identify threats. Toys can interact with their owners.


Local processing


The Neurala Brain and Brains for Bots software can run on the device (local or “at the edge”). Internet access is not required.


Products can act without waiting for a response delayed by Internet routing.  Privacy can be maintained.


High definition processing


The Neurala Brain and Brains for Bots Hosted provide additional processing power for more detailed analyses.


Videos from inspections and for ground truthing can be given extra attention for even better accuracy.


Fast pre-training


The Neurala Brain can learn objects and concepts with a handful of examples.


Get started quickly. Products can ship to end-users with knowledge built-in.


Incremental “on-the-fly” learning


The Neurala Brain can be updated in real-time with new information and without re-creating from scratch.


End users can customize The Neurala Brain for their specific needs.


Scalable


The Brains for Bots software runs on ARM, Intel and NVIDIA processors ranging from SBCs to servers.


Developers can select the hardware that is best for the product.  Applications can be written once and used everywhere.


Easy-to-use SDK


The Brains for Bots SDK contains an easy-to-use set of C ++ APls, sample code and documentation. Android and iOS wrappers simplify mobile development.


Applications can be written and deployed quickly without the need to know how to build deep learning neural networks.

Advantages of the Neurala Brain

FEATUREDESCRIPTIONBENEFIT

Real-time “what and where”

A product with Neurala Brain and the Brains for Bots SDK can find and recognize objects and concepts in real-time.

Drones can act while in flight. Cars can know what is nearby. Security systems can identify threats. Toys can interact with their owners.

Local processing

The Neurala Brain and Brains for Bots can run on the device (local or “at the edge”). Internet access is not required.

Products can act without waiting for a response delayed by Internet routing. Privacy can be maintained.

High definition processing

The Neurala Brain and Brains for Bots Server provide additional processing power for more detailed analyses.

Videos from inspections and for ground truthing can be given extra attention for even better accuracy.

Fast pre-training

A Neurala Brain can learn the objects and concepts important to your business with a handful of examples.

Products can ship to end-users with knowledge built-in.

Incremental “on-the-fly” learning

The Neurala Brain can be updated in real-time with new information and without re-creating from scratch.

End users can customize a Neurala Brain for their specific needs.

Scalable

Brains for Bots runs on ARM, Intel and NVIDIA processors ranging from SBCs to servers.

Developers can select the hardware that is best for the product. Applications can be written once and used everywhere.

Integrated functionality

Vision, audio, navigation, motor control, work together

Implementaton is fast.





The Neurala Brain in action

Real-Time Machine Learning and Visual Tracking

Collision Avoidance in Drones

Neurala Follow Autonomous Robot Control

Object Recognition and Classification

Deep Autonomy on GPUs

Neurala’s deep learning neural networks technology uses a proprietary set of algorithms to emulate, in software, the way biological brains control sensing, cognitive processing, spatial awareness, navigation, and motor control. 

A cornerstone of Neurala’s IP is a patent centered on running Artificial Neural Networks and Deep Networks on Graphic Processing Units, or GPUs (US8,648,867, “Graphic ProcessorBased Accelerator System and Method”). Deep (Neural) Networks are algorithms whose execution and training methods are highly conducive to parallelism. While CPU are designed to have a few cores optimized for sequential serial processing, GPUs have been built by design as massively parallel processors consisting of thousands of smaller cores which are best at handling multiple tasks simultaneously. 

Neurala’s patented and patent-pending technology leverages GPUs, when available, to emulate, in software, the way human eyes process information by moving a high-resolution “retina” around and quickly classifying objects. Neurala’s visual system is able to efficiently segment, classify, localize objects, and can be trained on demand and learn new objects within seconds. The vision system is connected to a navigation system that enables machines to self-localize without GPS, and avoid collisions based on anticipatory (not purely reactive) actions. Thanks to our patented GPU-friendly implementations, our algorithms can concurrently run in real-time, as a software-only solution. 

  • This field is for validation purposes and should be left unchanged.
×

Selfie Dronie Feature list

FLIGHT FEATURES

NEURALA INTELLIGENCE FEATURES

AUTOMATIC FEATURES

VIDEO AND SHARING FEATURES

SAFETY FEATURES

WARNINGS

DISPLAY

ALWAYS AVAILABLE HELP SYSTEM

MULTI-LINGUAL CONTROLS AND USER CONTROLS

×
  • “Can be trained by the user to recognize new types of objects in a matter of seconds”
    Xconomy
  • “Neurala’s real value-proposition is its ability to execute machine learning frameworks, locally, with limited resources.”
    TechCrunch
  • “Able to function without having to rely on the internet or communicate personal information to a third party”
    ZDnet
  • “Neurala is hardware agnostic, capable of scaling from low-end CPUs to high-end GPUs.”
    TechCrunch
  • “Neurala’s ‘Anywhere Brain’ means that devices can learn on the job”
    DroneLife
  • “useful for mobile applications such as autonomous vehicles that may not always have access to fast networks.”
    Machine Design