LILYGO Launches ESP32-S3-Based T-Camera S3 for TinyML Computer Vision Projects

LILYGO has launched a new Espressif ESP32-S3-based development board for computer vision at the edge and on-device tinyML workloads: the T-Camera S3, which includes display, camera sensor, microphone, and passive infrared (PIR) motion sensor.

Built around the Espressif ESP32-S3 system-on-chip, the T-Camera S3, brought to our attention by Linux Gizmos, has two 32-bit Tensilica Xtensa LX7 microprocessor cores running at up to 240MHz plus an ultra-low-power RISC-V coprocessor — added to the design shortly before Espressif's decision to go all-in on the free and open source RISC-V architecture for all future products. The chip includes 512kB of static RAM (SRAM) and 384kB of flash memory plus radios for 802.11b/g/n 2.4GHz Wi-Fi and Bluetooth 5 Low Energy (BLE) with Bluetooth Mesh support.

To this, LILYGO has added 8MB of pseudo-static RAM (PSRAM), 16MB of external flash, and an impressive array of on-board sensors: a microphone, a passive infrared (PIR) sensor for motion detection, and an OmniVision OV2640 image sensor offering a UXGA (1600×1200) maximum resolution at 15 frames per second, rising to 60 frames per second at CIF (352×240).

The camera's primary purpose is to feed data for tinyML machine vision projects, which can take advantage of the vector instructions included in the ESP32-S3 — boosting performance considerably over microcontrollers without native vector support. There's a JST connector for an optional battery, for on-the-go use, with a USB Type-C port for power and programming from either the Arduino IDE or VS Code. Finally, an on-board 0.96" OLED display, based on the SSD1306 controller, offers a 128×64 resolution local view for visualizations.

The company is now selling the T-Camera S3 on its AliExpress store at $21.83, or $24.68 with optional protective shell.