Seeed Studio Promises True No-Code Artificial Intelligence Development with SenseCraft AI
Supporting Chromium-based browsers, the new platform includes a 400-strong model zoo and rapid model training.
Seeed Studio has announced a new entry in its SenseCraft platform, SenseCraft AI — promising a beginner-friendly browser-based no-code development platform for those experimenting with artificial intelligence.
"Inspired by Google Teachable Machine and YOLO World," Seeed's Lily Li explains, "SenseCraft AI is a free, web-based platform designed for beginners, focuing on a no-code approach and an application-orientation. Our goal is to simplify and accelerate the creation of AI applications, making the process beginner-friendly and accessible to all. We've made everything web-based. No troublesome installation or configuration is needed. You can visit the SenseCraft AI platform from the [Google] Chrome or [Microsoft] Edge browser anywhere in the world."
There has never been more interest in machine learning and artificial intelligence, even as arguments surrounding the exact definition of both terms rage, and it's not all surrounding the large language models (LLMs) running on high-power cloud equipment in energy-guzzling data centers: there's plenty of light being shone on tiny machine learning (tinyML) running on resource-constrained microcontroller platforms, too.
It the latter end where SenseCraft AI focuses, reaching upwards into more powerful machine-learning-specific embedded hardware like the NVIDIA Jetson family too. The browser-based development platform is designed for use with six of Seeed's product lines at the time of writing, with more to follow: the XIAO ESP32S3 Sense, Grove Vision AI Module V2, SenseCAP Watcher, reCamera, recently-launched SenseCAP A1102 LoRaWAN Vision AI Sensor, and reComputer NVIDIA Jetson Edge devices.
The company promises a true "no-code" approach to model training and deployment, including a "beta feature" to create object detection models based on simply entering the name of the object with no additional training data. "While we're still improving the accuracy," Li admits, "you can enhance results by adding your own training images." Once trained, or selected from a 400-strong zoo of pre-trained models, and deployed, SenseCraft AI models can output their results over general-purpose input/output (GPIO) pins, UART, I2C, or SPI buses, or over a network via MQTT, Seeed has confirmed.
SenseCraft AI is now live on the Seeed Studio website.