Hillson Ghimire
I’m Hillson Ghimire, a geospatial data scientist specializing in remote sensing and machine learning. My work spans soil moisture and drought modeling, satellite-based disaster monitoring, and computer vision for agriculture.
I’m currently a PhD student at Rice University, advised by Noemi Vergopolan. My research explores how satellite observations, land surface models, and machine learning can be combined to estimate and predict soil moisture at high spatial resolution and better understand how agricultural drought develops.
Before Rice, I completed my MS at South Dakota State University, where I worked with Maitiniyazi Maimaitijiang. I developed deep learning models for wheat phenotyping and co-developed WheatAI, a web platform that helps plant breeders measure crop traits from imagery.
I was a research associate at the Geoinformatics Center at the Asian Institute of Technology, where I used satellite imagery to monitor natural hazards and their impacts for the Sentinel Asia initiative.
Email / CV / Google Scholar / GitHub / LinkedIn
Recent News
- Aug 2026Started my PhD in Earth, Environmental and Planetary Sciences at Rice University.
- Aug 2026Awarded the Ken Kennedy Institute Computational Science and Engineering Recruiting Fellowship ($15,000) at Rice University.
- Jul 2026Completed my MS at South Dakota State University. thesis
- Jun 2026Our wheat spike and spikelet detection paper was published in Scientific Reports. paper
- Jun 2026Served as Teaching Assistant and guest lecturer for Remote Sensing for Disaster Response at MIT Lincoln Laboratory, through August.
- Jan 2026Released WheatAI v1.0.
- Jan 2026Received the Dr. Fritz Gritzner Outstanding Graduate Geography Student Award at South Dakota State University.
- Oct 2025Received a USDA North Central SARE Graduate Student Grant ($19,999) to develop Wheat-TDM, a wheat tiller density mapping tool using multiscale remote sensing.
Publications
Ghimire et al. 2026, Fig. 3, CC BY-NC-ND 4.0
Wheat spike and spikelet detection and counting from high-resolution digital imagery using YOLO with oriented bounding boxes paper / data
H. Ghimire, M. Maimaitijiang, S. Thapa, S. Kaushal, D. J. Koupal, M. M. Billah, K. Poudel, S. Sehgal, M. Singh, S. Subedi, U. U. R. Janjua, and S. K. Sehgal, Scientific Reports, 16, 27355, 2026
We evaluate YOLOv11 and YOLOv12 for wheat spike and spikelet detection and counting with oriented bounding boxes, and introduce a benchmark dataset of 48,521 spike and 60,404 spikelet instances. YOLOv11 reached 95.8% mAP@0.5 for spike detection, with a counting correlation of r = 0.993, and 99.0% mAP@0.5 for spikelet detection.
M. Maimaitijiang, H. Ghimire, S. Thapa, M. M. Billah, S. Sehgal, M. Singh, S. Kaushal, K. Poudel, S. Subedi, U. U. R. Janjua, L.-O. Makonga, J. Halder, H. S. Gill, M. Sher, J. S. Sidhu, and S. K. Sehgal, arXiv:2601.08863, 2026
We present a browser-based platform that applies computer vision and deep learning to wheat phenotyping from smartphone, UAV, and portable microscope imagery. It supports spike and spikelet counting, Fusarium-damaged kernel estimation, kernel morphometrics, and automated stomatal analysis, with the aim of reducing labor costs and rater subjectivity in high-throughput selection.
Illustrative photo: Jacek Halicki, CC BY-SA 4.0, resized
Spectral–structural fusion for wheat lodging severity assessment using UAS imagery and deep learning preprint
S. Sehgal, M. Maimaitijiang, J. Halder, K. Poudel, U. U. R. Janjua, H. Ghimire, and S. Subedi, agriRxiv preprint, 2026
We combine UAS multispectral reflectance with photogrammetric canopy height in a six-channel dataset and train a 2D convolutional neural network to classify wheat lodging into four severity levels. The model reached 68.6% test accuracy, demonstrating the combined spectral–structural representation while highlighting the need for larger, more balanced datasets.
Ghimire et al. 2026, Fig. 5, CC BY-NC-ND 4.0
AI-assisted wheat spike and spikelet detection and density mapping using multiscale remote sensing data to support wheat grain yield forecasting thesis
H. Ghimire, MS Thesis, South Dakota State University, 2026
This thesis develops oriented object detection for wheat spikes and spikelets from close-range and UAV imagery. Across four YOLO-OBB architectures and seven UAV image modalities, image modality mattered more than model architecture. The trained models were integrated into the WheatAI platform.
Presentations
Dec 2025
Automated wheat spike detection and counting using UAV-based digital imagery and deep learning
Dec 2025
Predicting wheat grain yield and quality using multitemporal and multispectral UAS imagery with transformer models
Nov 2025
Integrating genomics and high-throughput phenotyping with machine learning for predictive breeding in winter wheat abstract
Nov 2025
Advancing wheat disease phenotyping through YOLOv11 and YOLOv12: Automated detection of FHB severity and damaged kernels abstract
Nov 2025
From image to insight: AI-driven wheat monitoring and yield prediction with multi-scale sensing abstract
Apr 2025
Deep learning-assisted wheat yield estimation through spike and spikelet counting from high-resolution imagery
Mar 2025
Deep learning-based detection and counting of wheat spikes and spikelets using high-resolution field imagery for improved yield estimation
Feb 2025
Wheat spike and spikelet detection on close-range digital imagery using deep learning abstract
Nov 2024
Wheat head detection from close-range digital imagery using deep learning photo
Oct 2024
Wheat head detection from close-range digital imagery using deep learning
Dec 2022
The intricacies in the implementation of integrated geospatial platform, database, and application for disaster risk management in Uttarakhand paper