ECEN 681 Seminar Spring 2026 Shima Salehi скачать в хорошем качестве

ECEN 681 Seminar Spring 2026 Shima Salehi

Title: "Multispectral Sparse-View 3D Scene Reconstruction and Structural Analysis for Complex Agricultural Environments" Abstract: Robust 3D reconstruction under sparse and spectrally diverse imaging conditions remains a fundamental challenge in Agricultural phenotyping. Standard reconstruction frameworks often struggle to preserve fine structural detail and global geometric consistency when limited viewpoints, complex plant morphology, and multispectral variability are present. In this seminar, I present a frequency-aware and multispectral extension of 3D Gaussian Splatting designed to address these limitations. The proposed framework, Local–Global Discrete Wavelet Transform 3D Gaussian Splatting (LGDWT-GS), introduces joint global and patch-wise wavelet supervision to regulate spatial-frequency learning during optimization. This approach stabilizes geometry while preserving high-frequency structural details under sparse-view conditions. The method is further extended to a multispectral setting, jointly reconstructing RGB and near-infrared modalities under shared geometry to ensure spectral and spatial coherence. To enable rigorous evaluation, we introduce an open-source few-shot 3D Gaussian Splatting benchmark along with a controlled multispectral greenhouse dataset spanning multiple crop species and spectral bands. Beyond controlled environments, I discuss the adaptation of this framework to field-based cotton imaging, where motion, illumination variation, and sparse viewpoints pose significant real-world challenges. Moving beyond surface reconstruction, I investigate structural topology through skeleton extraction and leverage foundation models such as Meta’s Segment Anything Model 3 (SAM3) and SAM3D to enhance geometric consistency. In parallel, I explore generative approaches for plant modeling, advancing toward semantically structured and controllable 3D plant representations. Biography: Shima Salehi is a Ph.D. student in computer engineering at Texas A&M University and a member of the Advanced Vision and Learning Lab (AVLL) under the supervision of Dr. Joshua Peeples. She received her B.S. degree in Electrical Engineering from Isfahan University of Technology, Iran, in 2020, and her M.S. degree in Computer Science from Amirkabir University of Technology (Tehran Polytechnic), Iran, in 2023. Her research focuses on developing novel computer vision and machine learning algorithms for 3D scene reconstruction, particularly frequency-aware 3D Gaussian Splatting for sparse-view and multispectral imaging, with applications in Agriculture. She has also worked on multimodal retrieval using large vision-language models and predictive modeling