Apple Plantation RGB-D Datasets with Feature-Based Image Selection (FBIS) for 3D Reconstruction in Agriculture

This Dataset is for the publication: “Feature-Based Image Selection (FBIS) for Enhancing 3D Reconstruction in Agriculture”.It contains RGB-D image sequences of apple plantations captured under two different seasonal and environmental conditions: Unbloomed dataset (before blooming, mainly trunks and branches) Bloomed dataset (during full bloom, dense foliage and occlusions) For each condition, we provide both the original dataset and the filtered dataset obtained using the FBIS algorithm, which reduces redundancy by selecting informative images for 3D scene reconstruction. Each dataset folder includes: RGB-D image streams (both RGB and Depth aligned at 640×480, 30 fps) Camera intrinsics (camera_intrinsics.json) Additionally, a separate folder contains COLMAP reconstruction videos generated from the FBIS-processed datasets (Bloomed + Unbloomed), illustrating the effect of the pre-processing step on 3D reconstruction quality. This dataset provides a benchmark for testing feature-based image selection, RGB-D reconstruction, and robotic vision in agricultural environments.