Week 17 : Wildcard Week (Machine Learning)¶
An AI recitation was conducted during Week 6; however, there was no opportunity to practically create a project on it. Therefore, this week is used to explore the integration of AI with physical objects and embedded system. Cursor was used as a collaborative AI assistant to accelerate the development process and support rapid prototyping of the embedded system workflow explored during this week.
Week 15 assignment could be categorized as follows :
- Project concept and objective
- Workflow and development process
- Tools, softwares and machines used
- Fabrication, assembly and final output
basics of AI
Project concept and objective¶
Summer is here and maintaining regular hydration becomes increasingly important. Therefore, this project explores a system that combines on-device computer vision and audio-based machine learning, designed to run locally without storing video data or identifying individuals
Concept: Hydrate is a small, dispenser-mounted assistant built around the XIAO ESP32-S3 Sense. It uses the onboard camera to detect when someone is in the drink zone and the microphone to recognize pouring sounds and simple voice commands such as “later.” When someone is present but has not had water for a set interval, it gives a brief audio and visual reminder; when a pour is detected or the user snoozes, it updates its state and stays quiet until the next cycle.
ok now i want to add a grove lora moule for 6 axis acceloremter & compass. to which gpio should i plut it inive: The goal is to encourage regular hydration at a shared water station without being intrusive. The system combines on-device computer vision and audio machine learning (TinyML) with simple rule-based logic so it can run locally without storing video or identifying people. The project also demonstrates a reproducible Fab workflow: CAD/CAM enclosure design, embedded firmware, trained edge models, and full documentation so others can rebuild and adapt it.
Workflow and development process¶
Library installation just follow the wiki
PSRAM (Pseudo Static RAM) is an external type of high-capacity memory that behaves like RAM but is built using DRAM technology with a built-in refresh controller. In simple terms, it gives small embedded devices extra working memory beyond their built-in SRAM.
What PSRAM does
Microcontrollers like the ESP32 have limited internal RAM (often ~320 KB usable SRAM). That’s not enough for heavy tasks like:
Camera image buffering Audio processing (TinyML models) Computer vision (frame storage) Running larger AI models or stacks
What is Edge AI?
Edge AI means running AI models directly on a device (“at the edge”) instead of sending data to the cloud.
Edge = local device
Examples:
ESP32-S3 Raspberry Pi Smart cameras Phones
What is Edge Impulse?
Edge Impulse is a development platform used to build and deploy machine learning models on small embedded devices like microcontrollers.
write about frame more frame bytes for noise what is it measured and also what
what does the max98357a do
convert voice process how to do it
upload image as jpeg
check labeling queue, the ai helps to save the process just ha
explain about espressif esp-eye
Processing Image → Image DSP (or “Image” processing block) Learning Image classification — MobileNetV2 96×96 or Transfer learning (Images) what is this
what does generate features do?
aim 85% is already okay
labeling
add picture at your phone
difference with sensecraft, tiny ml etc
difference between image detecgtion with boundinng boxes and classification
to include the model basically open the example and then add to then include into the file
need same position as where you want to put it or else...
what is opi psram, usb cdc on boot