Terminology
The table below defines key terms used in this document.
Term |
Definition |
|---|---|
Palette |
Sima’s suite of software tools for AI model development, video pipeline creation, SoC software, operating systems, and all software going into SiMa’s hardware |
Palette Edgematic |
Sima’s visual tool for evaluating AI/CV application deployment through SiMa’s hardware and software |
Model |
A software that can be trained to perform a cognitive task like a human being. For example, vision or speech analysis |
Application (pipeline) |
An application in AI/CV includes processes that need to be performed from sensor data ingestion to actual inference output. This involves pre-processing and post-processing of images and video streams |
Plugin |
A software or hardware component that can be part of an AI/CV application to accomplish a very specific task viz. H.264 standard based decoder |
Stream |
A continuous flow of audio/visual content. For AI/CV this often involves a video feed from a source such as a Real-Time Streaming Protocol (RTSP) server over Ethernet |
MLA |
SiMa’s machine learning accelerator block for running AI inferencing |
SoC |
System-on-Chip |
Allegro |
Block inside SiMa’s SoC responsible for video encoding/decoding |
EV74 |
Digital signal processor (DSP) block inside Sima’s SoC |
A65 |
ARM® processor responsible for running software and coordinating activities on EV74 and the MLA |