DOCSEARCHbyAlgolia
Submit Docs

pyWiFi ESP32-S3 Documentation

website·Indexed 32 days ago

https://wiki.01studio.cc/

Development guides and tutorials for the pyWiFi ESP32-S3 using MicroPython. Covers hardware interfacing including UART, I2C (OLED), ADC, PWM, and sensors (DHT11, DS18B20, MLX90614), as well as network communication via WiFi, Socket, MQTT, and BLE. Includes advanced features like AI vision for face and QR code recognition, Neopixel control, and REPL debugging.

Tokens
646.7K
Snippets
2.8K
Records
4K
Agent score
54%

What's inside pyWiFi ESP32-S3

  1. Overview of K210 KPU capabilities

    The KPU is a dedicated neural network processor within the K210 chip designed for low-power convolutional neural network (CNN) calculations. It enables real-time object detection and classification by identifying target size, coordinates, and category. Key technical specifications include:

    • Supports fixed-point models trained via mainstream frameworks (following specific constraints).
    • No direct limit on network layers; supports individual configuration of input/output channels and dimensions per layer.
    • Supports 1x1 and 3x3 convolution kernels.
    • Supports arbitrary activation functions.
    • Maximum model size for real-time operation: 5.5MiB to 5.9MiB.
    • Maximum model size for non-real-time operation: Flash capacity minus software size.

  2. Overview of AI Vision capabilities for ESP32-S3

    The ESP32-S3 MCU includes vector instructions designed to accelerate neural network computations and signal processing, enabling high-performance image recognition. Supported AI vision applications on the pyWiFi-ESP32-S3 platform include:

    • Color recognition
    • Motion detection
    • Cat face detection
    • Human face detection
    • Human face recognition

  3. Overview of 01Studio Online Model Training

    The 01Studio online training platform allows users to build YOLO11 visual models (Image Classification and Object Detection) without coding. The platform supports dragging and dropping images to build datasets and automatically generates model files and Python scripts for execution on development boards.
    Website: https://ai.01studio.cc

  4. Overview of pyBalance available resource packages

    The pyBalance resource downloads are categorized into the following packages:

    • Development Tools: Development software and necessary drivers.
    • Example Source Code: All source code associated with the online tutorials.
    • Firmware: 01Studio pyBalance MicroPython firmware, which includes factory-default example code.
    • Hardware Documentation: Development board schematics and interface description images.
    • Product Images: High-resolution photographs of the product.

  5. Overview of 01Studio Online Model Training

    The 01Studio online training platform allows users to build YOLO11 vision models without writing code. It supports image classification and image detection. Once training is complete, the platform automatically generates the model file (.kmodel) and Python scripts for deployment on supported hardware.

    • Website: https://ai.01studio.cc
    • Supported Platforms: 01Studio CanMV K230 / CanMV K230 mini
    • Requirement: Use firmware image version v1.4 or higher.

  6. Overview of the cv_lite image processing module for CanMV K230

    The cv_lite module is a lightweight image processing library for the CanMV K230, implemented using OpenCV. It serves as a high-performance supplement to the standard openmv image module, offering accelerated versions of common computer vision tasks. While openmv typically processes RGB565 images, cv_lite is optimized for RGB888 images and provides significant frame rate (fps) improvements for tasks such as blob detection, circle/rectangle detection, edge detection, and filtering.

  7. Overview of pyBalance MicroPython Robot

    pyBalance is an open-source self-balancing robot project by 01Studio designed for MicroPython. It allows developers to implement features such as self-balancing, forward/backward movement, line tracking, ultrasonic obstacle avoidance, and WiFi/Bluetooth remote control using the MicroPython language.

  8. Overview of cv_lite for CanMV K230

    cv_lite is a lightweight image processing module for CanMV K230 implemented using OpenCV. It is designed to provide accelerated versions of common image processing tasks to supplement the standard OpenMV image module, particularly for scenarios where OpenMV's performance is insufficient. While OpenMV typically processes RGB565 images, cv_lite is optimized for RGB888 images.

  9. Overview of the CanMV Open Source Project

    CanMV is an open-source project created and maintained by Canaan Technology. It ports MicroPython to the K210 edge computing AI chip, which features a 64-bit dual-core RISC-V architecture with a hardware FPU and a convolutional accelerator. The project enables developers familiar with Python or MicroPython to quickly implement machine vision, machine hearing, and deep learning applications on low-cost, scalable AI modules.

  10. Overview of the CanMV Project

    CanMV is an open-source project created and maintained by Canaan Technology. It ports MicroPython to the Canaan K210 and K230 edge computing AI chips. The K230 is a 64-bit dual-core RISC-V processor with a hardware FPU and convolutional accelerator, designed for low-cost, scalable AI modules for machine vision, machine hearing, and deep learning applications using Python.

  11. Overview of 01Studio AI Model Training for CanMV K230

    01Studio provides a no-code online training platform to build YOLO11 vision models. The platform allows users to upload images, build datasets, and generate model files and Python scripts for direct execution on development boards.

    • Online Training URL: https://ai.01studio.cc
    • Supported Hardware: 01Studio CanMV K230 / CanMV K230 mini
    • Requirement: Use image version v1.3 or higher.
    • Supported Vision Types: Image Classification and Image Detection.

  12. Overview of K210 KPU

    The KPU (Neural Network Processor) in the K210 is designed for low-power convolutional neural network (CNN) computations, enabling real-time object detection, classification, and face detection. It supports fixed-point models from mainstream training frameworks, 1x1 and 3x3 convolution kernels, and arbitrary activation functions.

    Memory constraints:

    • Real-time operation: Supports model parameters between 5.5MiB and 5.9MiB.
    • Non-real-time operation: Maximum parameter size is limited by (Flash capacity - software size).