clock menu more-arrow no yes

Filed under:

Movidius puts deep learning chip in a USB drive

New, 5 comments
The Fathom Neural Compute Stick (pictured without its casing)
The Fathom Neural Compute Stick (pictured without its casing)
Movidius

Today Silicon Valley chip maker Movidius released the Fathom Neural Compute Stick. It looks like a measly thumb drive, but inside it packs a high-end visual processing unit that can do a bunch of advanced image recognition. That chip, which is called the Myriad 2, is the same one powering the computer vision and autonomous features in DJI's latest drone. The Fathom is basically a plug-and-play version of the Myriad 2, and Movidius hopes engineers will use it to build deep learning features like like pixel-by-pixel imagine labeling and advanced video analytics into their existing products.

"It lets you implement machine learning in an ad hoc manner," Cormac Brick, head of machine learning at Movidius, tells The Verge. Brick sees applications for the Fathom in the development of drones, robotics, security, and virtual and augmented reality. Because it can plug into any device that has a USB port, developers don't need to redesign or prototype new products to roll out machine learning features.

The Fathom appears to be just another step in Movidius' mission to bring advanced computer vision to our devices. In January, Google announced it would be putting a version of the Myriad 2 in its next generation of Android handsets to support advanced image processing tasks like facial recognition. Currently, these kinds of features are limited to cloud-based apps, which rely on powerful servers to process and identify images. The Myriad 2 can perform these tasks locally, in real time, which is important for devices that need to react quickly like drones and robots. And with the Fathom, Movidius wants to give engineers a way to tap into these capabilities sooner than later.

"It lets you implement machine learning in an ad hoc manner."

Movidius hasn't released a final price for the Fathom but tells The Verge they're planning on offering it for under $100.