⁠Testing and coding the object detection model
Monitoring the each quadrant detection model on a computer screen using streamli
Assembling the main hardware component using tape and hot glue
Mounting the main hardware component on top a cardboard box to resemble a home system
Training and Labeling Object Detection AI Model
⁠Finalised prototype

SustAIn: Smarter Energy Management and ESG Reporting with AI

💡 Inspiration

Traditional motion sensors are highly inefficient, expensive to maintain, and often locked into proprietary ecosystems. We saw an opportunity to create a smarter, more cost-effective solution for energy management and ESG reporting.

🔍 What it does

SustAIn is an ULTRA low power, low-cost AI-driven zone management system. It uses custom-trained object detection to accurately monitor occupancy and control lighting. The system includes a smart dashboard for energy reporting and is built on an open-source cloud database.

🛠️ How we built it

We developed SustAIn using a combination of hardware and software:

AI on Edge Camera for object detection
ESP32 S3 microcontroller
8x8 LED Matrix and Neo Pixel LED for visual feedback
3D printed parts, including custom "human" models for training
Supabase for powering our Streamlit dashboard
AI IoT server to connect with the ESP32
Servo Motor to represent a Fan (Only turns on if Quadrant D is not 0)

🧗 Challenges we ran into

Where to even start 😭🔫

We fried 3 ESP32 microcontrollers due to bad circuit connections
Bad mobile data reception plagued our Supabase Cloud Database connections, leading to erratic behaviour for controlling the lights
Had to train AI model 5 times 😭😭 (but finally got the accuracy up to 94%)
our Arduino IDE just died 💀💀 cause of bad imports

🏆 Accomplishments that we're proud of

Developed a system where one Edge AI Camera does the work of multiple IR sensors
Achieved accurate ESG Reporting with Real Time Dashboard using Streamlit
Fixed all issues. YIPPEEE!
Designed an ultra-low power system (0.5W per camera)