Introduction to the ESP32-S31 with Vision

Welcome to the Introduction to the ESP32-S31 with Vision workshop!

About this workshop

#

In this workshop, you will explore the ESP32-S31 system-on-chip (SoC) and build computer-vision applications that run entirely on the device. You will use the ESP32-S31-Korvo-1 multimedia development board, its Board Support Package (BSP), ESP-Vision, and ESP-DL.

The workshop begins with the SoC and development board overviews. You’ll first use the ESP-VISION workflow to build camera and safety-zone applications in MicroPython—a lightweight implementation of Python designed for microcontrollers. Your Python code will run directly on ESP32-S31, rather than on your computer.

Then, you’ll contrast the ESP-VISION workflow with the traditional ESP-IDF workflow by using the ESP32-S31-Korvo-1 BSP and adding ESP-DL face detection to its camera example.

If you choose so, you will be able to use AI with the ESP-Vision MCP server for agent-assisted application development.

Inference remains local throughout the workshop. Camera frames do not need to be sent to a cloud service.

By the end of the workshop, you will know how to:

• Identify the processing, connectivity, multimedia, and security capabilities of the ESP32-S31.
• Use ESP-Vision and develop camera applications in Python.
• (optional) Connect an AI coding assistant to the ESP-Vision MCP server.
• Select an appropriate local model for an embedded vision task.
• Detect and track hands in camera frames.
• Detect when a hand enters a safety zone and generate a local warning.
• Use the ESP32-S31-Korvo-1 camera, display, touch, LED, audio, and storage hardware through its BSP.
• Load an ESP-DL model and run face detection from an ESP-IDF application.

Prerequisites

#

Required hardware

#

• Computer running Linux, Windows, or macOS.
• ESP32-S31-Korvo-1.
• USB Type-C data cable to connect your computer to the board’s UART port.
• USB Type-A to Type-A data cable to connect the ESP32-S31-Korvo-1 to the ESP-Vision Web IDE.
• (optional) microSD card formatted as FAT32 for storing model files that are too large for the ESP-Vision internal flash filesystem.
• (optional) Another USB Type-C cable for the dedicated power port.

Required software

#

ESP-Vision

#

• A desktop browser with Web Serial support for the ESP-Vision Web IDE and web flasher: Google Chrome 89+, Microsoft Edge 89+, Opera 76+, or Firefox 151+. Other Chromium-based browsers, such as Brave, may also work. Safari is not supported; Chrome or Edge is recommended.
• Cursor for the agent-assisted exercises. Other MCP-compatible agents can also be used, including Visual Studio Code with GitHub Copilot agent mode, Claude Code, and Claude Desktop.

ESP-DL and native ESP-IDF development

#

• ESP-IDF v6.1 or later; v6.1 is recommended. ESP32-S31 is a preview target in v6.1, so idf.py commands need the --preview option
• git and a serial terminal
• Optional: Visual Studio Code with the ESP-IDF extension

Agenda

#

• Overview of the ESP32-S31 SoC
• Overview of the ESP32-S31-Korvo-1 devboard
• Tooling: Getting started with ESP-Vision
• Tooling: Set up the ESP-Vision MCP server (optional)
• Assignment 1: LED blink with ESP-Vision
• Assignment 2: Your first camera application
• Lecture 1: Available models for ESP-Vision
• Assignment 3: Hand detection with ESPDet
• Assignment 4: Add a safety zone
• Assignment 5: Keep hands off!
• Assignment 6: Get started with the ESP32-S31-Korvo-1 BSP
• Assignment 7: Face detection with ESP-DL
• Conclusion

Time requirements

#

Next step

#

Overview of the ESP32-S31

Feedback

#

If you have any feedback about this workshop, feel free to start a new discussion on GitHub.