With 12 patents awarded, Eta Compute has a strong history of innovation and expertise in low power design and milliwatt machine learning. We have applied this expertise to the Machine Vision problem.
We achieve very low power intelligent vision by a combination of technologies and design decisions. Compared to traditional machine vision:
- We transmit only metadata, no full video streaming, a tremendous power saving
- We design specific μW vision sensors
- We use the most efficient available edge-AI inference processors
- We design specific neural networks and algorithms for tiny machines
- Our patented Tensai Flow tools deliver the best ML optimization for the hardware
Flexible Ultra-Low-Power Intelligent Vision Platform
We have a unique software-reconfigurable platform for ultra-low power intelligent vision. There are five elements of our “secret sauce”:
- Application specific dataset and data acquisition process
- Library of proprietary TinyML models and algorithms
- Tensai™ Flow AI compiler
- Validation and optimization framework
- Design methodology for ultra-low-power endpoint
Secure and Private
Our technology inherently protects privacy, no image ever leaving the sensor. The image processing and recognition is done on device, and the image is discarded afterward. Only meta information is shared on the network.
This greatly reduces the attack surface, and we implement state of the art security measures to ensure the highest standard of protection against various threats.