AI vision sensor camera learns through clicks

By Carolyn Mathas, contributing writer

DFRobot has announced its HuskyLens, a user-friendly, artificial-intelligence (AI) vision sensor featuring built-in powerful algorithms. The pre-installed algorithms enable facial recognition, object recognition, object tracking, line following, color detection, and tag detection. Faces can be detected at 30 frames per second.

Switching between the various algorithms is easy and is based on just one click. Users don’t need to understand the algorithms involved, making it a time-saving device for experienced makers as well as a powerful tool for hobbyists.

The performance of the HuskyLens is extremely high and it is suited for capturing quickly moving objects.

Features also include the ability for line following and tag recognition to be simultaneously activated and a 2.0-inch display that eliminates the need to connect to a PC for monitoring.

The HuskeyLens AI vision sensor camera improves through repetition, learning new objects, faces, and colors with a single click. It can learn the target from different perspectives and ranges by just holding the button down. HuskyLens is compatible with microcontrollers such as Arduino, Raspberry Pi, LattePanda, and micro:bit.

HuskyLens incorporates an AI-specialized chip Kendryte K210 and is powered by a 400-MHz dual-core RISC-V 64-bit processor, increasing the performance of the AI chip to be 1,000× faster, compared with the STM32H743 when running a neural-network algorithm.

The vision sensor enables users to unleash creative projects without knowledge of the complex underlying technology. HuskyLens is now available for backing on Kickstarter.