In-Sensor AI
Create unique AI-based features and embed them directly into smart sensors
Free Unlimited Plan

Neuton.AI enables the creation of ultra-small neural networks that can be embedded into resource-constrained smart sensors.

This opens up new possibilities for innovative device architecture

Benefits of In-Sensor AI

Reduced battery consumption: 
no need to wake-up the host MCU to read & process sensor data, all ML analysis inside the low-power sensor
Unloads MCU bandwidth:  
free the CPU from algorithm processing and thread switching, which is critical for hard real-time systems
Simplifies software development:
distributed systems are easier to design, you need to spend less time on multithreading and managing shared MCU resources

New Era of In-Sensor AI

Key pillars of success
  Neuton Neural Network Framework  
Automatic design of extremely compact and accurate models without additional compression
Ultra-low power MCUs and sensors
Ultra-low power sensors with built-in AI operate at the microwatt level

In-Sensor AI in Action

Smart ring as a Remote Control

Tiniest Gesture Recognition Model in ISPU
With this solution, you can use a smart ring to control your home electronics through various gestures:
Swipe Right
Next channel
Double Tap
Swipe Left
Prev channel
Rotation Clockwise
Volume up
IDLE – no gestures
Rotation Counter Clockwise
Volume down
Unknown gestures
7 >96%
Program RAM (kB) Data RAM (kB)
Total solution 10.7 1.4
Model 1.4 0.00
DSP 4.8 0.2
Inference engine 4.5 1.2
Estimated on ISPU LSM6DSO16IS

Touch-free Solution for Interacting with Smartwatches

Embedded within the Sensor is an AI Feature, Similar to that of the Apple Watch
The solution allows you to control a smartwatch by recognizing the following gestures:
Activate device
Double pinch
Double clench
5 >94%
Program RAM (kB) Data RAM (kB)
Total solution 3.2 1.4
Model 0.7 0.00
DSP 1.6 0.2
Inference engine 1.0 1.3
Estimated on ISPU LSM6DSO16IS

Teeth-brushing Tracking Solution

More insightful analysis for users about their teeth-brushing process
  • Real-time identification of 15 zones of the oral cavity
  • Incredibly small total footprint
  • On-device implementation
17 >95%
Program RAM (kB) Data RAM (kB)
Total solution 6.4 1.4
Model 3.1 0.00
DSP 2.3 1.2
Inference engine 1.0 0.2
Estimated on Cortex M-4

Logistics: On-device Package Tracking

The solution allows you to analyze the following box states directly on the device:
Free fall
Wrong orientation
Transported by courier
Transported by car
IDLE by car
7 >97%
Program RAM (kB) Data RAM (kB)
Total solution 9.2 0.2
Model 0.6 0.00
DSP 4.1 0.02
Inference engine 4.5 0.18
Estimated on ISPU LSM6DSO16IS

Daily Human Activities Recognition

Recognition of complex and similar human actions directly on the ISPU
Hand washing
Teeth brushing
Hand still
Face washing
Hair brushing
Unknown class
134 7 >94%
Program RAM (kB) Data RAM (kB)
Total solution 9.79 1.2
Model 0.65 0.00
DSP 4.3 0.14
Inference engine 4.83 1.06
Estimated on ISPU LSM6DSO16IS

Solutions by Leveraging Neuton.AI

gesture recognition
vital sign determination
human-machine Interface
human activity recognition
machine fault classification
asset tracking and monitoring

3 simple steps to make smart sensors AI-Driven

Collect data
Upload data and build automatically tiniest neural networks
Download compiled  library and run inference on your device
Sign Up
Infrastructure Credits
Use Neuton's free Zero Gravity plan, accompanied by eligible free trial credits to cover infrastructure costs.
Register as a new customer within the Google Cloud Platform to be eligible for up to $300 in free trial credits.
Corporate customers are also eligible for an additional $200 in credits on top of the $300 free trial credits. In order to be qualified to redeem the additional $200 credits, customers must register as a new customer with Neuton using a corporate email domain (no personal email accounts allowed, e.g. Gmail, Yahoo).
Once the customer has exceeded the available credits, infrastructure fees will then be changed until the subscription is terminated. The subscription may be canceled at any time by cancelling the pricing plan.
Google Cloud Platform Costs:
To build models with your data, an IT infrastructure is required. Google Cloud Platform (GCP) uses virtual machines. Every model can be trained in parallel without losing speed. That is why each dataset training requires a separate virtual machine with enough GPU. The platform automatically chooses and provisions the right infrastructure for your dataset to ensure fast learning.
The costs are calculated as follows:
Storage - $0.04 GB/month
Training - The cost per model for the training of one dataset depends on the number of rows in it and varies from $0.88-4.82
Training Costs per dataset size, per hour of training, per model
0-1000 rows of data - $0.88
1001-5000 rows of data - $1.44
5001-50,000 rows of data - $2.56
>50,000 rows of data - $4.82
For new accounts, GCP provides up to $500 credits, for the next 90 days. $500 is enough for at least 100 hours of training on the platform!
Stay updated, join the community