DEEPX will showcase its neural network computational processing solution at the the 2022 Embedded Vision Summit.
DEEPX, an AI semiconductor startup based on NPU (Neural Processing Unit) technology, will present its neural network computational processing solution, a comprehensive technology of hardware and software, at the upcoming 2022 Embedded Vision Summit in Santa Clara, California.
DEEPX will be introducing the DEEPX series (DX-L1, DX-L2, DX-M1, and DX-H1), an SoC product line with the world’s highest performance NPU in energy efficiency and processing performance that can compute cutting-edge AI algorithms with high AI accuracy. The DEEPX series specializes in more efficiently processing DNN algorithms for object recognition, face recognition, sound recognition, image classification, image enhancement, and more. The scalability of NPU technology can range from 0.1 TOPS for ultra-small sensors to tens of POPS for AI servers.
Consisting of four solutions of DX-L1, DX-L2, DX-M1, and DX-H1, the DEEPX series will be launched sequentially from the second half of this year to the first half of next year. These products will be delivered to more than 20 global customer companies.
At the exhibition, buyers will be able to see the performance of DEEPX’s FPGA-based NPU prototype through a live demonstration, and experience AI solutions for object detection, face recognition, image classification, and more first-hand should they visit the booth.
DEEPX is a startup that has entered the rapidly growing AI semiconductor market and researched and developed source technology for semiconductors with the world’s highest power performance ratio. DEEPX boasts implementation at the highest levels globally in 3 important technology categories of the AI semiconductor field: support for the latest AI algorithms, the world best AI accuracy, and the world’s highest ratio of computational performance to power consumption.
In the four years since its inception, DEEPX has achieved great results, including a cumulative investment of $25 million, the title of organizer of the development of Korea’s next-generation intelligent semiconductor, and 101 patent applications at home and abroad.
DEEPX succeeded in implementing the NPU applicable to IoT and edge devices for the first time in Korea in 2019, and from the following year, it developed technology to simplify the deep learning algorithm and optimized the hardware structure to upgrade the NPU design dozens of times. At long last, DEEPX will be releasing the DEEPX series this year.