I'm Eros.
DLSU CSE Undergraduate and Machine Learning Engineer for DLSU BCRU. Former Research Assistant under BLAST.
DLSU CSE Undergraduate and Machine Learning Engineer for DLSU BCRU. Former Research Assistant under BLAST.
I lead the data pre-processing, deep learning architecture training, hyperparameters optimization, and post-detection image analysis and processing. I also developed the model training to algorithm testing pipeline of the project. Our team developed an object detection and image processing system using embedded systems to monitor pests in cacao fruits. This automation improves pest management by reducing detection time and improving accuracy. Deploying the system in embedded systems also reduces power consumption and material cost. Publication
I lead the data pre-processing, segmentation architecture training and testing, and image post-processing. Our team developed a comprehensive system that combines computer vision with the Internet of Things paradigm for indoor plant monitoring. The system captures plant images and relevant environmental data at regular intervals to monitor crop health and how it is affected by external variables. The project has the potential to help farm managers reduce risks and economic loss in indoor crop production. Publication
I contributed in the dataset processing, augmentation, and model training. Our team addressed the object imbalance in deep learning detectors by generating synthetic Cocoa Pod Borer images using an auxiliary classifier generative adversarial network (AC-GAN). This model achieved a Frechet inception distance score of 7.14 for synthetic images, leading to an object detector with an average precision of 0.94, surpassing the baseline model's 0.88 Publication
I spearheaded data pre-processing, augmentation, model training, testing, and optimization, covering both object detection and time-series models. I also took charge of post-processing and analysis for both image and time-series data. Our undergraduate thesis received the Gold - Most Outstanding Thesis Award, the highest accolade in our department. We developed a comprehensive solution integrating sensor and image data for agricultural monitoring, specifically focusing on cherry tomatoes. his holistic system combines deep learning and time-series models to automatically detect crop phenotypes and predict future environmental variables. Moreover, the system is deployed on a low-cost embedded platform capable of autonomously collecting crucial data for analysis. This data is made accessible to end-users through a web application, providing valuable insights for crop management. Abstract