Combines Mindtech's Chameleon AI Training Platform with
GIGABYTE Technology Co. Ltd.'s servers enabling rapid creation of
massive datasets for training neural networks
LONDON, June 30, 2020 /PRNewswire/ -- Training AI
for visual tasks requires vast amounts of accurately annotated
images or video. The Chameleon platform creates and "films" virtual
worlds to produce synthetic data for neural network training;
synthetic data which accurately models real world appearances,
physics and behaviors. This is efficiently achieved at server
scale, and GIGABYTE's servers are designed with this type of task
in mind.
The Chameleon platform is designed to scale through effective
use of CPU(s) threads and cores, offloading to GPU(s) where
possible to fully utilize all available resources. The Docker
containerization of the platform ensures that running on the
GIGABYTE platform is effective and straightforward. This
server-scalability is used to allow Mindtech to create
"ready-to-train" Data Packs, where images and advanced annotations
are supplied, enabling customers to create AI assisted applications
in Smart City, Smart Machine and Smart Vision including retail,
premise management, robotics and transportation. These Data
Packs allow customers a rapid route to market, covering key data
requirements including difficult to capture corner cases.
"Creating and processing data for training AI takes significant
compute resource, and GIGABYTE's servers are an excellent platform
upon which to build a solution for synthetic data creation," said
Mindtech's VP of Product Management Chris Longstaff. "They offer an
industry leading balance of CPU, GPU and memory, in a power
efficient system making them an ideal partner for our Chameleon
platform. Close coupling the creation of the synthetic data
on the same GIGABYTE system that will be used for training the AI
system offers significant performance and efficiency gains."
"Successful implementation of neural networks requires huge
amounts of data," said GIGABYTE's VP of Engineering Andy Chen.
"Mindtech's Chameleon platform democratizes the availability of
this 'big-data' by allowing any company to create massive amounts
of perfectly annotated synthetic data. Through the
containerization of the Chameleon platform with technologies such
as Docker and Kubernetes, it is a perfect match to our DNN
platform, enabling rapid creation of large synthetic data sets
which in turn permits our customers to create innovative solutions,
with a fast time to market."
Mindtech's Chameleon platform and Data Packs are available for
immediate licensing.
For more information on Mindtech's Chameleon Platform please
contact chris.longstaff@mindtech.global.
For more information on GIGABYTE's servers please contact
server.grp@gigabyte.com.
Logo -
https://mma.prnewswire.com/media/971869/Mindtech_Logo.jpg
Logo
- https://mma.prnewswire.com/media/1196672/GIGABYTE_Logo.jpg