This project aims to develop a gas and smell detection system using the BME688 Devkit and an ESP32 microcontroller by applying ADAM optimizer machine learning algorithm on gas resistance data. The system mimics human smell capabilities and can detect various gases, making it ideal for industrial and robotics applications.
- BME688 Devkit: Environmental sensor for detecting temperature, pressure, humidity, and gas using AI.
- Adafruit HUZZAH32 ESP32 Feather: Handles sensor data, processes it, and sends real-time results via Bluetooth or stores it for later analysis.
- BME AI-Studio Desktop: A desktop tool for sensor configuration, model training, and analysis.
- BME AI-Studio Mobile App: For real-time data monitoring and labeling.
- BSEC2 Library: Allows custom code development using trained models.
- BME688 Development Kit Software: Firmware, which needs to be flashed to the ESP32 microcontroller by running the Flash.bat script in the downloaded files from the provided link.
- Board Configuration: Flash firmware, configure sensors. The config used for this project is available in the Src folder.
- Sample Collection: Gather environmental samples using the devkit and mobile app.
- Model Training: Use the BME AI-Studio Desktop to train models on collected data. The project file for the AI-Stdio Desktop are available here. Please unzip the files and open the project folder from the AI-Studio Desktop application to view my collected specimens and trained algorithms.
- Testing: Evaluate model performance in real-world conditions using the mobile app.
The system was able to distinguish different smells, such as air, coffee, and tea, achieving 86.66% accuracy. The classification results are depicted in this video.
Potential applications include:
- Air Quality Monitoring for smart cities.
- Health Applications in wearable devices.
- Smart Agriculture and Home Automation.
For more details, check out the Documentation folder.