Hackster's June Impact Spotlights Bring Imaginative Edge AI Solutions to Real-World Challenges

Hackster's June Impact Spotlights featured a trio of innovative community members who have leveraged edge AI to build projects that address real-world challenges. From traffic lights that recognize when you're wearing a helmet, to CNC mills that can detect failure before it occurs, and security cameras that keep watch without utilizing the cloud, each application highlights how far AI at the edge has evolved and how accessible it's becoming for makers and engineers everywhere.

No Helmet, No Green AI Traffic Light

First guest speaker, Roni Bandini, demonstrated his No Helmet No Green AI Traffic Light project that employs AI to detect if motorcyclists are wearing a helmet while stopped at traffic lights. If not, the light will remain red until the rider dons a helmet. Bandini took inspiration from Honda's No Helmet/No Green smart traffic light initiative that rolled out in Argentina in 2021.

Bandini's platform is designed around a Texas Instruments AM62A board outfitted with a USB camera, which he paired with Edge Impulse's machine learning algorithm that he trained (using a fun LEGO-based setup) to detect helmets with an accuracy of 95%. Bandini also built a server to pass detection results to a DFRobot UNIHIKER display, which acts as the light head, complete with custom Python scripts, a PHP server backend, and a 3D-printed mount.

Edge AI for Predictive Maintenance in Tormach CNC

Next speaker, Ajith K J with Tiny Prism Labs, showcased how to leverage edge AI for predictive maintenance in a Tormach CNC machine, which detects early signs of CNC milling failures via analyzing vibrations in real-time. The platform was developed using a Seeed Studio XIAO nRF52840 Sense board, equipped with an onboard six-axis IMU sensor mounted directly on the Tormach CNC mill to gather vibration data.

The analysis process works in two phases, with the first collecting vibration profiles across spindle speeds and milling patterns, and extracting statistical features, such as standard deviation, skewness, and kurtosis, which are used to train a TensorFlow Lite anomaly detection model. The model is then deployed on the XIAO board, allowing it to identify early warning signs of milling failure, which is processed locally on the edge.

The process allowed Ajith to reduce latency, increase data security, and eliminate the need for cloud access.

AI Security Camera for Vehicles

The third and final presenter, Solomon Githu, talked about his AI security camera for vehicles project that takes advantage of edge AI to identify humans who try to vandalize or attempt vehicle theft. Githu designed his AI security camera around an Arduino Potenta H7 and a Portenta Vision Shield to monitor vehicles and anyone who gets near them.

Githu simulated real-world situations, like tire theft and window smashing, then trained a custom image-classification model using those sounds and images. The Portenta runs his custom model at around one frame per second, and when the platform detects a threat, it automatically sends an alert via built-in Wi-Fi, triggering an email notification.

Conclusion

Whether it's road safety, machine maintenance, or vehicle protection, these projects demonstrate that there's no need for massive budgets or corporate backing to develop innovative AI-driven solutions for real-world needs. With open hardware, ingenuity, smart design, and a little imagination, real-world impact is just a prototype away.