Here we demonstrate that an image sensor can itself constitute an ANN that can simultaneously sense and process optical images without latency. Our device is based on a reconfgurable two-dimensional (2D) semiconductor photodiode array, and the synaptic weights of the network are stored in a continuously tunable photoresponsivity matrix. We demonstrate both supervised and unsupervised learning and train the sensor to classify and encode images that are optically projected onto the chip with a throughput of 20 million bins per second.
Running on servers in SK Telecom’s data center,
Alveo U250 cards have demonstrated improved
throughput and increased accuracy for Tview theft
detection services, enabling the company to provide
more customers with access to real-time, AI-based
security services, and a more reliable defense against
security threats.
Alveo U250 cards have demonstrated improved
throughput and increased accuracy for Tview theft
detection services, enabling the company to provide
more customers with access to real-time, AI-based
security services, and a more reliable defense against
security threats.
SiMa.ai is a new company developing a product to fit these requirements. Its MLSoC provides a custom ML accelerator that is rated at 10x better power efficiency than leading GPU-based designs and 2-3x better efficiency than the best accelerators available today. The MLSoC also includes a popular ARM-compatible CPU for legacy code and standard DRAM, Ethernet, and camera interfaces to simplify system design. Companies developing camera-based edge systems should consider the MLSoC for adding intelligence to their next design.