Projects, Fall 25


Embedded Machine Learning, ECE, Carnegie Mellon University
  • Wildfire Detector

    Early Detection of Wild Fires Using Drones and Machine Learning

    By Tanisha Sethi, Mario Cruz, and Erin Isabel Anand
    Post thumbnail
    Post thumbnail
    In order to combat wildfires we explore deploying autonomous drones with fire detection capabilities to enhance early identification and situational awareness. Some key advantages are: (1) Overcoming terrain barriers: Drones can navigate rugged, inaccessible regions where ground sensors or cameras are ineffective; (2) Enhanced Detection Resolution: Flying at lower altitudes... [Read More]
  • Trash Sorter

    A Smart Device to Help Sort Trash into Bins

    By Kejia Hu, Bowei Li, and Zijia Liu
    Post thumbnail
    Post thumbnail
    Our objectives and motivation are to build a smart device that helps people put trash in the right bin. The target users can be the general public using public recycle bins, organizations running sustainability or recycling programs, or people who have difficulty manually sorting trash. [Read More]
  • Traffic Sign Alert

    On-Edge Stop Sign Recognition and Alert System

    By Balaji Sathyanarayanan
    Post thumbnail
    Post thumbnail
    I enjoy driving my car both in India and in Pittsburgh. However, I have occasionally experienced challenges in detecting and responding to stop signs in a timely manner. This experience has motivated me to design a Stop Sign Recognition and Alert System. The goal of this system is to enhance... [Read More]
  • VoiceMod

    Fun Speech Converter

    By Alejandro Miranda
    Post thumbnail
    Post thumbnail
    ● What: Real-time speech-to-speech voice conversion system running on embedded hardware ● Why: Privacy-preserving, low-latency voice transformation for entertainment purposes ● For Whom: Entertainment/Recreational Users ● Product-type: Proof-of-concept Get the complete PDF report. [Read More]
  • Pet Robot

    A Fun Companion on your Desk

    By Mahlet Mesfin
    Post thumbnail
    Post thumbnail
    I wanted to create a responsive pet robot that provides light entertainment and interaction without increasing screen time. It is primarily for anyone seeking fun, interactive desk gadget. It reduces network usage because all inference happens locally; only occasional the model updates or telemetry need network transfer. Because it is... [Read More]
  • AegisEye

    Intelligent Pan-Tilt Face Tracking Camera

    By Jesse Barkley and Christian Shimp
    Post thumbnail
    Post thumbnail
    Our goal was to deploy a computer vision model on a lightweight, low energy, edge device to detect human faces. Being able to accurately identify human faces and track with low power devices has a wide range of applications from security to personal home robotics that require face identification. [Read More]
  • Offline Storytelling Companion

    Bedtime Story Companion for Kids

    By Shreesh Tripathi and Vignesh Natarajan
    Post thumbnail
    Post thumbnail
    What: An offline bedtime-story companion on Jetson Orin Nano that generates and narrates kid-safe stories using a tiny LLM + STT/TTS. It is a new product built from open-source components (LLM, TTS) with a safety-first pipeline. Our objective is to deliver safe, customizable, internet-free storytelling with less than 5s start-to-speech... [Read More]
  • Wearable Motion and Position Monitor

    For Sign Language Interpretation

    By Adwoa Asare, Tianyi He, and Shao-Ju Wang
    Post thumbnail
    Post thumbnail
    We wanted to create a hand position/movement classification glove with applications in human- computer interaction. This is useful for cases where hands need to be tracked without the use of a camera. We use edge processing to maintain user bio-data private and real-time responsiveness. We use multi-sensor fusion to improve... [Read More]
  • Guitar Helper

    Learn to Play the Guitar from other Songs

    By George Pan, Harvey Ko, and Harrison Lo
    Post thumbnail
    Post thumbnail
    Guitar Helper: an embedded ML system that listens to short audio snippets of strummed chords and classifies them in real time. Beginners struggle with chord memorization and fast chord changes. It is also difficult to find chord progressions online for less popular songs. We wanted to create a palm-sized “Guitar... [Read More]
  • Footstep Classifier

    Identify People from their Footsteps

    By Andrew Yu
    Post thumbnail
    Post thumbnail
    The objective of this project us using off-body sensors to classify footsteps and identify people. The motivation was that it is a cool project to try. The target use case is security. The embedded ML rationale for it was low cost, low latency, and a private solution [Read More]
  • Off the Cuff

    Blood Pressure Monitoring Device

    By Om Kulkarni, Pradhumna Guru Prasad, and Aidan Vogt
    Post thumbnail
    Post thumbnail
    In the current state, blood pressure monitoring is episodic and intrusive (cuff). The gap: Hypertension is dynamic – A snapshot once in a while misses the full picture. Our solution: An embedded deep learning enabled device with the capability of continuous, passive monitoring using a single optical sensor. [Read More]
  • ALFORM

    Articulatory Labeling For Optimal Real-time MRI Experiments

    By Joseph Konan
    Post thumbnail
    Post thumbnail
    The objective for this project is to build a device for MRI technologists and researchers who need reliable articulatory data without wasting scanner time. The ALFORM system IS A real-time articulatory tracking tool that shows which articulators are being used and how often, while you scan. Unlike expensive trial-and-error experiments... [Read More]
  • Mind Drive

    EEG-Controlled Arduino Car

    By Runtian (Toby) Yang, Patarapornkan (Pat) Anantarangsi, and Yicheng Rong
    Post thumbnail
    Post thumbnail
    Electroencephalogram (EEG), also known as “brain wave”, is a promising brain computer interface (BCI) technology that is low-cost and noninvasive. Our goal is to build a prototype device that can read EEG, decode the user intent, and use that to control an Arduino Car. The target users would be disabled... [Read More]
  • CatMotion

    Cat Behavior Classification

    By Lin Zhan, Junrui Zhao
    Post thumbnail
    Post thumbnail
    Our Goal is to develop a wearable device that classifies cat behaviors (e.g., rest, walk, feed) using motion sensor data as a cat-mounted smart collar system. Our motivation is to catch early changes in cat behavior that may signal cat health issues, and our objective to enable continuous, real-time cat... [Read More]