Electronics Weekly
Called RZ/V2N, it comes with the company’s own AI accelerator, DRP-AI3.
The 15Top/s (INT8) claim comes from sparse networks through the accelerator’s “advanced pruning technology”, said the company.
On dense networks it can only hit 4Top/s, and its Resnet50 score is 769 inference/s.
The main CPU is a 1.8GHz quad-core Arm Cortex-M55, backed by a single core 200MHz Cortex-M33.
Read all the latest news from Embedded World 2025 »
There is four lane x two channel MIPI-CSI connectivity, H.265 and H.264 video codec support, and an Arm Mali-G31 GPU.
Dual MIPI channels allow two cameras to capture a single scene from two directions.
“The two-camera system significantly improves spatial recognition performance compared to a single camera, and enables more precise human motion line analysis and fall detection,” claimed Renesas. “Furthermore, the dual-camera system captures images from different locations, allowing a single chip to count cars in a parking lot and recognise license plates.”
AI vision applications are expected – for traffic and congestion analysis in commercial facilities, industrial for visual inspection and driver monitoring, for example.
See a live demonstration on Renesas’s stand (234) in hall 1 at embedded World this week, or find RZ/V2N on this Renesas web page
