Sponsored by Infineon, The Goting Machine Learning Contest featured an innovative edge machine learning application.
Participants were challenged to use DeepCraft Studio PSOC 6 AI Kit in the creation of ML projects for smart home and industrial applications. We had to build edge AI models by collecting sensor data that utilizes the onboard sensors in the kit, introducing models that work in the real world, and documenting solutions as hackstar tutorials.
After reviewing all submissions, here are eight projects that emerged as winners:
Detection of AI mental abnormalities
The heart is the most important muscle in the body, and heart damage is the main cause of death, but for a variety of reasons, not everyone keeps it in great condition. More people need regular cardiac tests to reduce the risk of the heart. Testing often requires bulky equipment in expensive hospitals. So, Edriver and Victor Alta Mirrano Developed Wearable Electrical Devices Use the Dry electrodes and SparkFun AD8232 Heart Rate Monitor.
Dried electrodes are allowed to use by athletes for electrical measurements during activity. Using machine learning, the collected data is organized and presented to caregivers in ways that lead to actionable insights to address the issue of over-information from wearable health devices.
Please note that this project is for “empirical and descriptive purposes only” and should not be used in place of certified medical devices.
Maintenance of AI-equipped predictive industrial machinery
Predictive maintenance employs data and analysis to reduce the risk of equipment failure and failure, extend equipment lifespan and reduce the need for switchgears.
Eslam is fade''' project is a Predictive Maintenance Model For industrial machinery. Used to collect real-time vibration data. Training data for the model was collected using the PSOC 6 Integrated Inertial Measurement Unit (IMU), trained in Deepcraft Studio, and deployed to the PSOC 6 via Modustoolbox.
Fayed tests his model with a vacuum cleaner, but the project can be adapted to other industrial machines.
SmartListener: Ambient Sound Classifier
Sound classification uses waveform characteristics to group audio signals from applications such as voice recognition, music genre detection, and security monitoring into predefined categories.
Mohammed Arivedea I've created it SmartListener To classify sounds in home environments. This model was trained using audio samples from the ESC-50 (Environmental Sound Classification) dataset, converted to a 16kHz bitrate for compatibility with DeepCraft Studio.
There are five separate classes for sound classification, including baby crying, fire alarm, glass breaking, footsteps, and bark. This device listens to these specific sounds and sends an MQTT alert when they are detected. Using a Tensorflow Lite model trained in Deepcraft Studio, there is a 3D printed enclosure to deploy in any room.
Smart Scale: Generates freshness detection
Smart Scale Hardware Architecture
Thousands of completely edible fruits and vegetables are disposed of each day. Because they don't meet cosmetic standards. Picture-Fectect produce is an unrealistic standard that leads to large quantities of food waste. Some supermarkets are working to reduce this waste by donating surplus food and selling “ugly” fruits and vegetables at discounted prices.
Milan Fels Comero's Smart Scale The aim is to introduce objective standards for “evaluating and selling fresh produce.” The exterior doesn't tell a complete story, so radar imaging and AI are used to “assess the true internal and external state of fruits and vegetables.”
This serves as part of a retail checkout experience where customers place fruit items on smart scales, and devices use built-in neural networks to calculate freshness scores. SmartScale applies proportional discounts based on the score and prints the customer's price labels. This reduces waste and allows customers to make sustainable decisions.
It currently works with bananas, but can be expanded to more fruits and vegetables.
Theia: Tracking Heuristic Evaluation Intelligence Analyzer
Situational awareness is an important part of home security and automation. Brian Staley and Brayden's project uses PSOC 6 radar sensors and edge AI models Analyzes entrance and exit traffic and classifies the objects that pass through.
This project consists of three separate applications for training, inference and processing. TheIA-REC RECORDS and THEAIA-INF perform model inference, and TheIA-Client processes the results and displays information to the user.
Due to time constraints, the model has two main classes: people and balls. It can detect people and balls coming and going into the field of view of the radar sensor.
AI Blender Speed Detector
Guillermo Perez Guillens Projects will be detected Current status of the blender It uses the PSOC 6 on-board accelerometer and provides corresponding audio or LED alerts. Detects when the blender is running. This idea can be applied to any device that operates using a DC motor in any setting.
The boards had to be positioned in different directions to capture at least two axes of the accelerometer. This project supports LEDs and voice alerts via the EDU DFR0699 Voice Recorder module.
Illegal logging detector
Illegal logging is a major problem in most developing countries. For example, the Congo Basin loses 1-5% of its forest cover each year due to illegal logging, and the majority of the wood consumed in Mexico is of illegal origin.
Alejandro Sanchez'device Detect chainsaw sounds and human voices in protected areasindicates that illegal logging may occur. The PSOC 6's onboard microphone gets audio signals that the model classifies into one of three classes: Chansawi, People, or Forest (base state).
Intellifaan
There's nothing like a cool breeze blowing in the summer heat. Most fans are blown in a single direction and need to be adjusted manually. Wouldn't it be cool if your fans chase you around the room rather than the road? That's the reason Litick and Jill Construction of De Anza College Intellifan, Fans that automatically track people and respond to hand gestures.
This project uses the gesture model of the prerequisites of Deepcraft and the PSOC 6 radar sensor to recognize five different swipe and push gestures. The web interface displays a live camera view and provides speed, tracking, and manual positioning controls. The team plans to upgrade their current products with voice commands, batteries for portable use, and tower designs for better airflow.
