technical Projects

Breadboard Model of the GreenSat ICU

GreenSat: On-Board Image Classification Unit

GreenSat is an on-board image classification unit for use in the Maya-7 nanosatellite to be launched in December 2025. It utilizes artificial intelligence in high-performance, low-power microcontrollers of STM32. It is designed to receive raw images from the satellite's camera and classify images as agricultural land or non-agricultural land. The classifications and metadata are expected to be transmitted to the Philippine Space Agency's (PhilSA) ground station in UP Diliman for the duration of the satellite's life. This is an ongoing project done in collaboration with De La Salle University and the Philippine Space Agency.

Keywords: STM32, Microcontroller, Artificial Intelligence, Image Processing

Evaluation of Quantized CNN Architectures for Land Use Classification for Onboard Cube Satellite Computing

This is a research paper evaluating three different pre-trained models and the effects of quantization on these models using Quantization Aware Training (QAT) in PyTorch. This was evaluated using the EuroSAT RGB Dataset for Land Use Image Classification

Published paper: IEEE, Original File

Keywords: CNN, Quantization, Land cover classification, Transfer learning

Quantized Model Structure

TinkerCAD Circuit Simulation

Plant.tio

Plan.tio is a scalable and modular automated plant-watering system that can maintain and monitor the health of multiple indoor potted plants. This utilizes one master board and multiple slave boards, one for each plant.

Project files: Github

Keywords: Arduino, C++, TinkerCAD

Land Cover Mapping of Boracay Island, Philippines Using Remote Sensing and Machine Learning

This is a research project that aimed to compare the accuracy and performance of three machine learning models: Gaussian Mixture Model, Random Forest, and K-Nearest Neighbor algorithms. These models were used to classify an image of Boracay Island, masking white pixels as built-up areas and black pixels as non-built-up areas.

Published paper: IEEE, Original File

Keywords: QGIS, Remote Sensing, Machine Learning

Original image

Built-up and non-built up areas

Application of Different CNN Models for Land-Use and Land-Cover Classification

Collaborators: Julian Ducut

This is a research project that aimed to compare different AI models for image segmentation in land-cover mapping. This compared the performances of three AI models: ResNet50, VGG16, EfficientNet-B0.

Published paper: IEEE, Original File, Github

Keywords: Python, Remote Sensing, Artificial Intelligence, Image Segmentation

WounDAR: LiDAR and Machine Vision based Wound Assessment

Collaborators: Reign Balajadia, Mary Cotoco, Lance Lim, Alvin Chua Jr.

This was a capstone project that aimed to perform wound analysis for remote healthcare. It utilizes LiDAR and machine vision to capture images, measure, and analyze the wound including predicting likelihood of infection. The resulting data is summarized and can be presented to a doctor for diagnosis and monitoring of slow-healing wounds.

Published paper: IEEE, Original File, Github

Keywords: Python, Machine Vision, Artificial Intelligence

Wound Classification (NSFW)

Computer Vision Module

Pneumatic Simulation

Soft Gripper Design for Plastic Waste

Robotic Handling Operation

Collaborators: Reign Balajadia, Mary Cotoco, Lance Lim, Alvin Chua Jr.

This was a thesis project that aimed to classify and detect plastic waste for a six-armed soft robotic gripper made of silicone to approach and grab the object for trash sorting. The image to the left is the computer vision approach of object detection, then classification to determine appropriate action through selective actuation of the six-armed soft gripper powered by pneumatics.

Project files: Github

Keywords: Python, Computer Vision, Artificial Intelligence

Philippine Junior Data Science Challenge

Collaborators: John Vincent Cortez, Lance Garcia, JR Salonga Beo

This project is an exploratory data analysis of a movie dataset provided by the organizers. The insights on the dataset are shown via static visualization as seen in the PDF file to the right. This was analyzed using Python and visualized using Photoshop.

Project files: Github

Keywords: Python, Data Science

EDA1.pdf

Data Model for a Ground Receiving Station

Ground Receiving Station Database Structure

This project is a proposed database structure for a ground receiving station in the Philippines. It is designed to manage and store satellite images from various satellites and easy retrieval of satellite images for end users.

Project files: Github

Keywords: Python, Firebase, NoSQL, Database

Vehicle Image Classification for Parking Structures

This project is a vehicle image classification tool using artificial intelligence to determine if the vehicle is a car, motorcycle, or bicycle. Each class of vehicle has its respective parking rate and is calculated accordingly using this program.

Project files: Github

Keywords: Python, Artificial Intelligence, Image Classification

Vehicle Log File

MMDA Twitter Timeline Dataset

Collaborators: Hazel Kate Companero, Roque Perez, Marife Villalon

This project is an exploratory data analysis of the MMDA Twitter timeline dataset. The dataset was scraped and analyzed for insights on the frequency and location of vehicular accidents. The image to the right showcases the number of accidents based on location plotted onto the map of Metro Manila.

Project files: Github

Keywords: Python, Data Science

Financial Inclusion in ASEAN

Collaborators: Sabina Ma, Rina Sagrit, Cielo Vicencio

This project is an exploratory data analysis and data visualization of the financial inclusion dataset by the World Bank. This analysis focuses on the ASEAN region and is visualized in a locally hosted web server.

Project files: Github

Keywords: Python, JavaScript, HTML/CSS, Data Science