Sipeed Launches MAIX-II Development Kit for Deep Learning

Following the successful launch of the K210-based MAIX-I series AIoT boards, Sipeed has announced the release of the MAIX-II designed around the popular Allwinner V831 camera processor. The dev kit is aimed at smart home, medical, industrials, and even agricultural applications like pest and agricultural monitoring.

"The MAIX-II core module is an entry-level edge computing (built-in NPU) board that can run in a Linux environment," which means we can expect MAIX III, IV, V to be more powerful yet compact with larger memory to run bigger models. There was a need to change the SoC as the K210 processor came with small RAM that had made it difficult to run complex models, and it couldn't support a high-resolution camera.

So, when MAIX-II is compared with its predecessor, it's equipped with higher CPU clock frequency, larger memory (RAM), and more peripherals — that includes an Ethernet port, ADC, and audio port. Since MAIX-I did not support video encoder, this board features H.264 up to 1080p at 30fps, H265 up to 1080p at 30fps, and JPEG up to 1080p at 30fps.

Sipeed has provided an improved SiP 64MB DDR2 along with optional 16M flash, while the AI accelerator is capable of 0.2TOPS NPU for facial recognition and object recognition, etc. The recommended power supply is about 5V/1A through any Type-C interface.

Sipeed adapts the Jupyter Notebook development environment for MAIX-II, making it easier for embedded AI developers to run Python code. For those interested, the MAIX-II Dock Kit — consisting of a MAIX-II core module and MAIX-II Dock baseboard — is available for pre-order via Seeed Studio for $28.80 and is estimated to be shipped somewhere around January 15th.

Also, speaking about the boards that are expected to be out soon. The MAIX-III is said to be capable of 1.5TOPS NPU (vip9000), featuring dual-band Wi-Fi, Gigabit Ethernet, and approximately 0.5~2GB DDR3 RAM and priced in the $15-$30 range.