LILYGO T-Bao

Version History:

Purchase Links

Table of Contents

• Description
• Preview
• Modules
• Quick Start
• Pin Overview
• Related Tests
• FAQ
• Projects
• Resources
• Dependent Libraries

Description

T-Bao is an open-source AIoT terminal based on ESP32-S3-WROOM-1 and K210 RISC-V AI chip, featuring high performance, low power consumption, expandability, and programmability. It deeply integrates K210 RISC-V architecture AI chip with ESP32 dual-core Wi-Fi/BLE communication module, equipped with professional-grade OV2640 2-megapixel camera (supporting 180° rotation adjustment) and ST7789V 1.54-inch IPS touch screen (90° adaptive rotation display), achieving millisecond-level face recognition and dynamic image processing.

Built-in MAX98357A I2S digital power amplifier, MSM261S high-sensitivity microphone array and hardware codec unit, supporting voice interaction and high-definition audio output. Uses AXP202 intelligent power management system, supports 3.7V lithium battery power supply and ultra-low power operation, integrates MPU6050 six-axis sensor, TF card storage expansion and rich interface resources (UART/SPI/I2C/Touch), fully compatible with Arduino/MicroPython ecosystem. Suitable for industrial vision detection, smart security, AIoT terminals and educational robots.

Preview

Physical Image

Pin Diagram

Important Notes

⚠️ SD Card Usage Notes

Modules

AI Processor (K210)

• Chip: Kendryte K210
• Architecture: RISC-V Dual Core 64bit with FPU
• Frequency: 400MHz (can be overclocked to 600MHz)
• Memory: 8MB SRAM
• NPU: KPU Neural Network Processor
• Features: Supports YOLOv3, Mobilenetv2, TinyYOLOv2, face recognition and other models

Main Controller

• Chip: ESP32-D0WDQ6-V3
• Core: Xtensa LX6 Dual Core
• FLASH: 16MB
• PSRAM: 8MB
• Clock: 240MHz
• Wireless: Wi-Fi 802.11 b/g/n, Bluetooth 4.2 + BLE

Display

• Size: 1.54-inch IPS
• Resolution: 240x240px
• Display Type: IPS LCD
• Driver Chip: ST7789V
• Touch: Capacitive Touch Screen
• UI Framework: LVGL

Camera

• Sensor: OV2640
• Pixels: 2 megapixels
• Features: Supports 180° rotation adjustment

Audio

• Amplifier: MAX98357A I2S
• Microphone: MSM261S High Sensitivity Microphone Array

Sensors

• Motion Sensor: MPU6050 (Six-axis gyroscope + accelerometer)
• Power Management: AXP202 PMU

Motor Driver

• Chip: DRV8833

Overview

Quick Start

Example Applications

K210 Firmware Flashing

1. Download the latest firmware: MaixPy Firmware
2. Download flashing tool: kflash_gui
3. Select corresponding firmware and model files for flashing

Note: Speech recognition application requires writing the model file maix_asr_2900k_0x500000.kmodel to address 0x500000 first, then flashing the firmware.

Model Training

Use Maixhub to train your own models.

Maixhub provides model training and model sharing functions. You only need to prepare the dataset for training, without setting up training environment and code. Upload training data to quickly train models, making it convenient to quickly create your AI applications, or to learn the process and principles of AI training.

Additionally, you can download your favorite models from the MaixHub model library, and even upload and share models you created yourself, whether they are trained on MaixHub or various interesting models trained with your own code!

Access Maixhub

1. After registering and logging into your account, go to the official website and click on Model Training

2. Click Create New Model, and enter the name and select the model type

3. Select dataset, click Upload Dataset

Here you can choose your own dataset, or select a dataset already prepared by Maixhub, or upload your own dataset.

Note: When uploading datasets, please ensure the dataset format is .zip file, and after decompression, the dataset contains a folder named dataset, which contains the training dataset.

4. Select training parameters, click Start Training

Note: The selection of training parameters needs to be adjusted according to your own requirements, such as training epochs, learning rate, optimizer, etc.
!! The deployment platform currently supported by our examples for K210 platform is nncase, and the trained model file suffix is .kmodel.

5. Wait for training to complete, download model file

6. After downloading, upload the model file to the T-Bao device, and upload the main.py and .kmodel files from the compressed package to the SD card

You can use MaixPy IDE for simulation running, or use MicroPython for running.

Development Platforms

1. Arduino IDE - ESP32 Development
2. MaixPy IDE - K210 Development
3. ESP-IDF - ESP32 Advanced Development
4. MicroPython - Script Development

Related Tests

Test data to be completed

FAQ

• Q. How do K210 and ESP32 communicate?
  A. Communication is through UART serial port, with K210 as AI processor and ESP32 as main controller and communication module.

• Q. Camera not working properly?
  A. Check camera ribbon cable connection, ensure camera module is properly inserted, check power supply.

• Q. Speech recognition inaccurate?
  A. Ensure use in quiet environment, check if microphone is blocked, try retraining the model.

• Q. How to extend battery life?
  A. Use AXP202 power management functions, turn off camera and screen when not needed, enable low power mode.

• Q. SD card not recognized?
  A. Refer to SD card installation diagram in Important Notes, ensure card is properly inserted and formatted as FAT32.

Projects

• T-Bao Schematic
• Example Code

Resources

• Schematic
• K210 Datasheet
• ESP32-S3 Datasheet
• OV2640 Datasheet
• ST7789 Datasheet
• MicroPython Documentation

Dependent Libraries

• MaixPy - K210 Development Framework
• LVGL - Embedded Graphics Library
• Arduino_GFX - Graphics Display Library
• ESP32 Arduino Core - ESP32 Arduino Support
• Adafruit_Sensor - Sensor Library