Seeed Studio Teases the Ultra-Compact Edge AI TinyML XIAO ESP32S3 Sense Smart Camera Dev Board

Seeed Studio has shown off a work-in-progress board that aims to shrink edge vision applications down as small as possible: the XIAO ESP32S3 Sense, with a modular camera expansion board for minimal footprint.

Based, as the name implies, on the Espressif ESP32-S3 system-on-chip (SOC), the XIAO ESP32S3 includes two 32-bit Tensilica Xtensa LX7 cores running at up to 240MHz, 384kB of static RAM (SRAM), and 512kB of on-chip flash plus a real-time clock block with a RISC-V-based ultra-low-power coprocessor. A radio block offers 2.4GHz single-band 802.11b/g/n Wi-Fi connectivity along with Bluetooth 5.0 Low Energy (BLE) connectivity.

To this Seeed has added 8MB of pseudo-static RAM (PSRAM) and 8MB of off-chip flash storage, while still cramming the resulting board into its compact XIAO form factor. Unlike previous entries in the range, though, it's not a single-board solution but one split into two.

A factory-fitted expansion board includes an Omnivision OV2640 camera module, offering a 1632×1232 (UXGA) image sensor running at up to 15 frames per second — or 60 frames per second, if you drop down to CIF resolution. The design of the board makes the camera removable, Seeed notes, meaning dropping in a more powerful sensor like the OV5640 will be possible post-purchase.

The board also includes an unspecified "high-quality" microphone and a microSD card slot for storage, with Seeed positioning the part as ideal for tinyML edge AI work with audio and video data — aided by the ESP32-S3's inclusion of vector instructions, designed to accelerate machine learning workloads.

Although Seeed is showing off functional prototypes of the board, product manager Chunchun Tian tells us that the XIAO ESP32S3 Sense is "still under development" with a release date penciled in for April this year; pricing has not yet been confirmed.