This research paper presents a workflow for 3D reconstruction of archaeological objects and contexts using low-cost UAVs and image segmentation techniques. The goal is to automate the process of generating dense point clouds from photographs, reducing the need for expensive and specialized hardware. The workflow consists of three main phases: generating binary masks to select regions of interest, producing dense point clouds, and transforming them into triangular meshes. The study uses a low-cost UAV (SG906 Pro) and free, open-source software (OpenSfM, OpenCV, and MeshLab) to capture and process data from 151 images of three prehispanic monoliths. The paper highlights the importance of binary masks in improving the visual quality and detail of the reconstructed models. The results show that using binary masks significantly reduces noise and enhances the accuracy of the 3D reconstructions. However, establishing quantitative frameworks to evaluate the quality of the generated models remains challenging due to the irregular surfaces and recent discovery of the objects. The study concludes with recommendations for future improvements, including the need for larger and more diverse datasets to increase the detail level of the point clouds and triangular meshes.This research paper presents a workflow for 3D reconstruction of archaeological objects and contexts using low-cost UAVs and image segmentation techniques. The goal is to automate the process of generating dense point clouds from photographs, reducing the need for expensive and specialized hardware. The workflow consists of three main phases: generating binary masks to select regions of interest, producing dense point clouds, and transforming them into triangular meshes. The study uses a low-cost UAV (SG906 Pro) and free, open-source software (OpenSfM, OpenCV, and MeshLab) to capture and process data from 151 images of three prehispanic monoliths. The paper highlights the importance of binary masks in improving the visual quality and detail of the reconstructed models. The results show that using binary masks significantly reduces noise and enhances the accuracy of the 3D reconstructions. However, establishing quantitative frameworks to evaluate the quality of the generated models remains challenging due to the irregular surfaces and recent discovery of the objects. The study concludes with recommendations for future improvements, including the need for larger and more diverse datasets to increase the detail level of the point clouds and triangular meshes.