The field of 3D reconstruction and mapping is rapidly advancing, with a focus on developing innovative methods for environmental monitoring and archaeology. Recent developments have seen the integration of UAV technology, photogrammetry, and machine learning to improve the accuracy and efficiency of 3D modeling and segmentation. Notably, researchers are exploring the use of self-supervised learning and semi-supervised approaches to address challenges such as limited labeled data and visual occlusion. These advancements have significant implications for applications such as crop monitoring, infrastructure management, and archaeological site discovery.

Some noteworthy papers in this area include: Iterative Motion Compensation for Canonical 3D Reconstruction from UAV Plant Images Captured in Windy Conditions, which presents a pipeline for generating high-quality 3D reconstructions of individual agricultural plants. Demeter: A Parametric Model of Crop Plant Morphology from the Real World, which introduces a data-driven parametric model for encoding plant morphology. From Pixels to People: Satellite-Based Mapping and Quantification of Riverbank Erosion and Lost Villages in Bangladesh, which demonstrates the use of a general-purpose vision model for tracking riverbank erosion and lost villages. Self-supervised Pre-training for Mapping of Archaeological Stone Wall in Historic Landscapes Using High-Resolution DEM Derivatives, which proposes a segmentation framework for automatic mapping of low-lying dry-stone walls using high-resolution DEM derivatives.