I am a PhD student in the Computer Vision Lab at UMass Amherst, advised by Prof. Subhransu Maji. My research explores the relationships between computer vision tasks, with a focus on multi-task and transfer learning, and developing robust embeddings for large-scale scientific applications. My primary application domain is environmental remote sensing, where I build models to analyze multispectral satellite imagery for hydrology.

Before my PhD, I was a graduate research assistant at the University of the Philippines working on 3D reconstruction under the supervision of Prof. Rowel Atienza, where I completed my M.S. in Electrical Engineering. I obtained my B.S. in Electronics and Communications Engineering from the same univeristy.

News

• 2025-12 Attending AGU’25! I’ll be giving two oral presentations on our latest work in remote sensing: RiverScope and WildSAT
• 2025-11 SuperRivolution accepted to WACV 2026!
• 2025-11 RiverScope accepted to AAAI 2026! I will also be giving an oral presentation at AAAI.
• 2025-06 WildSAT accepted to ICCV 2025! Also presenting it at the Computer Vision for Ecology workshop as a Spotlight talk
• 2025-03 Our paper on Improving Satellite Imagery Masking using Multi-task and Transfer Learning has been published at IEEE JSTARS
• 2025-01 Gave a talk at the University of the Philippines about “Computer Vision in the Wild”
• 2024-05 Passed PhD Portfolio with Distinction! (Awarded to select PhD students meeting a high standard of completion, voted by faculty)
• 2024-03 Task2Box paper accepted at CVPR 2024 as a Highlight (11.9% of accepted papers)!
• 2023-10 COSE paper presented in VIPriors at ICCV 2023
• 2022-09 Started PhD CS at University of Massachusetts Amherst
• 2020-06 REIN accepted (oral presentation) at CVPR 2020 workshop
• 2018-01 Started MS Electrical Engineering at University of the Philippines

Selected Publications

(* means equal contribution)

	SuperRivolution: Fine-Scale Rivers from Coarse Temporal Satellite Imagery Rangel Daroya, Subhransu Maji WACV 2026 (to appear) arXiv We introduce a method and an accompanying dataset that leverages multiple low-resolution satellite imagery instead of expensive high-resolution imagery for fine-scale satellite image tasks such as river width estimation
	RiverScope: High-Resolution River Masking Dataset Rangel Daroya, Taylor Rowley, Jonathan Flores, Elisa Friedmann, Fiona Bennitt, Heejin An, Travis Simmons, Marissa Jean Hughes, Camryn L Kluetmeier, Solomon Kica, J Daniel Vélez, Sarah E Esenther, Thomas E Howard, Yanqi Ye, Audrey Turcotte, Colin Gleason, Subhransu Maji AAAI 2026 (to appear) (Oral Presentation) arXiv We introduce a high-resolution river segmentation dataset that can be used for precise hydrology tasks such as river segmentation and river width estimation
	WildSAT: Learning Satellite Image Representations from Wildlife Observations Rangel Daroya, Elijah Cole, Oisin Mac Aodha, Grant Van Horn, Subhransu Maji ICCV 2025 CV4Ecology Workshop @ ICCV 2025 (Spotlight Talk) ICCV25 / arXiv We show satellite image representations can be improved using wildlife observations
	Improving Satellite Imagery Masking using Multi-task and Transfer Learning Rangel Daroya, Luisa Vieira Lucchese, Travis Simmons, Punwath Prum, Tamlin Pavelsky, John Gardner, Colin J Gleason, Subhransu Maji IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 2025 JSTARS25 / arXiv We show a multi-task model for predicting water, cloud, cloud shadow, terain shadow, and snow/ice simultaneously from satellite images
	Task2Box: Box Embeddings for Modeling Asymmetric Task Relationships Rangel Daroya, Aaron Sun, Subhransu Maji CVPR 2024 (Highlight: 11.9% of accepted papers) project page / CVPR24 / arXiv We present a method for modeling asymmetric relationships between vision tasks using box embeddings
	COSE: A Consistency-Sensitivity Metric for Saliency on Image Classification Rangel Daroya*, Aaron Sun*, Subhransu Maji ICCVW 2023 project page / ICCVW23 / arXiv We propose metrics related to the consistency and sensitivity of saliency maps on image classification tasks to assess the performance of different explainable AI methods on a variety of models and datasets
	REIN: Flexible mesh generation from point clouds Rangel Daroya, Rowel Atienza, Rhandley Cajote CVPRW 2020 CVPRW20 We present a method for generating a mesh from a sparse point cloud with an arbitrary number of points