Computer Vision Research Intern Agricultural AI

Subin Kim

*********@***.*** 404-***-**** GitHub LinkedIn Logan, Utah, USA

Professional Summary

Computer vision and agricultural AI researcher with experience in UAV hyperspectral imaging, crop disease monitoring, deep learning model training, and sensor-based perception. Published first-author ICPR 2026 work on spectral-spatial fusion for wheat rust detection. Strong background in PyTorch, TensorFlow, OpenCV, Linux, MATLAB, data mining, Dataset-Adaptive Model Design, and AI system architecture. Education

Utah State University – M.S. in Computer Science (GPA: 3.70 / 4.00) Logan, UT Thesis Defense Completed Aug 2024 – Aug 2026

Research Focus: Computer Vision, Machine Learning, Deep Learning, Agricultural AI, Sensor Perception

Gyeongsang National University – B.S. in Mechanical Engineering Jinju, South Korea Core Foundations: Dynamics, Robotics, Mechanical Measurements, CAD, Manufacturing Mar 2017 – Aug 2023 Technical Skills

Languages & Tools Python, Bash, Git, Linux (Ubuntu), Conda, MATLAB, ROS ML & Deep LearningPyTorch, TensorFlow, Keras, Scikit-Learn, Timm, Hugging Face Computer Vision OpenCV, UAV/Drone Imagery, Hyperspectral Imaging, Remote Sensing, Image Classification, 3D Medical Imaging

Agricultural AI Crop Disease Detection, Crop Monitoring, NDVI-based Vegetation Filtering, Plant Physiological Knowledge, Dataset-Adaptive Model Design, Unsupervised Learning Data & Systems Data Analysis, Data Mining, Basic SQL, Feature Extraction, Model Training, Experimental Evalu- ation, AI System Architecture

Peer-Reviewed Publications & Patents

1. Subin Kim, Xiaojun Qi. “Dual-Branch Spectral-Spatial Network with Knowledge- and Data-Driven Band Selection for UAV Hyperspectral Wheat Rust Detection.” International Conference on Pattern Recognition (ICPR), 2026. (Accepted, First Author). [Code]

2. H. Lim, D. Jeong, S. Kim, et al. “Novel Concept of Liquid Level Sensor using Distributed Optical Fiber Sensor with High Spatial Resolution.” KCI Spring Symposium (2022) & KR Patent App. KR10-2023-0192798 (2023). Research Experience

Agricultural Computer Vision for UAV Hyperspectral Wheat Rust Detection May 2025 – Apr 2026 Utah State University – Computer Vision Lab (Advisor: Dr. Xiaojun Qi) Logan, UT

• Designed and implemented an end-to-end agricultural computer vision pipeline for UAV hyperspectral crop disease monitoring under noisy field conditions, resulting in an accepted ICPR 2026 first-author publication.

• Trained PyTorch-based deep learning models for crop disease classification by integrating spatial image features with compact spectral representations from UAV hyperspectral imagery.

• Reduced spectral dimensionality by 93.6% from 125 bands to 8 optimized VNIR bands, improving storage and computational efficiency for resource-constrained agricultural AI workflows.

• Achieved 85.19% Accuracy and 85.72% Macro-F1, outperforming heavy full-band Transformer and MambaHSI baselines by 4.6 percentage points.

• Built an NDVI-based vegetation masking and data refinement pipeline to reduce label noise and spatial mismatch in agricultural field imagery.

Robotics & Autonomous Systems Researcher Sep 2022 – Dec 2023 Gyeongsang National University – Safe Search Lab, Intelligence and Interactive Robotics Lab Jinju, South Korea

• Won Second Prize (Sonnet.ai Proprietor Award) at the 1st F1Tenth Korea Championship (2022), competing in real-time autonomous racing systems.

• Tested, debugged, and tuned a 1/10-scale autonomous vehicle with LiDAR perception and motion planning algorithms in a Linux/ROS environment.

• Contributed to a PPO-based reinforcement learning controller for OpenSim musculoskeletal models, experimenting with simulation-based control policy design.

Selected Technical Projects

3D CT Spine Fracture Detection Pipeline (Medical Imaging) [Code]

• Architected a 3D medical imaging ETL pipeline for the 343.5 GB RSNA 2022 dataset; processed DICOM images, segmentations, and bounding boxes via HU windowing and isotropic resampling for a 3D ResNet-18 baseline. Lightweight Deepfake Detector (Computer Vision & Signal Processing) [Code]

• Built a signal-processing and data-mining based ensemble classifier on the 40 GB FaceForensics++ and CIDAUT datasets using LBP, wavelet, and color statistical features; achieved a 0.71 F1-score and 0.77 Accuracy. AI System Architecture Course Project

• Designed modular AI system workflows with clear pipeline boundaries, data contracts, failure analysis, and evaluation- oriented documentation for end-to-end AI application development.

Contact this candidate