Linearity Multi-View RGB Dataset of Wheat Ears and Zamia Plant for 3D Reconstruction with Segmentation Annotations

This dataset is primarily designed to support research in multi-view 3D reconstruction of plant structures using RGB imagery. All data were acquired using a standard RGB smartphone camera in both field and laboratory conditions. The dataset provides circular and multi-angle image capture of wheat ears and whole plants, enabling applications such as structure-from-motion (SfM), neural radiance fields (NeRF), and other 3D reconstruction pipelines. The dataset is organized into three multi-view sequences and three additional data folders, one containing videos, and the other two containing annotated image sets.

Dataset structure:

Sequence 1–3: Wheat Ears in Field Conditions

Each sequence contains RGB images of wheat ears captured outdoors in field conditions. The wheat ears are positioned inside a rectangular frame to spatially constrain the scene. Images were taken from multiple viewpoints around the frame, providing multi-angle coverage suitable for 3D reconstruction. These sequences reflect natural illumination and real-world environmental variability.

Zamija: Laboratory Videos of a Zamia Plant 

This folder contains four RGB videos of a Zamia plant recorded indoors in a controlled laboratory setting. Each video captures a full 360° circular motion around the plant, enabling frame extraction for 3D reconstruction workflows. The videos provide dense viewpoint coverage under artificial indoor lighting conditions.

wheat_wo_background_w_labels: Dried Wheat (White Background, Annotated Images)

This folder contains RGB images of dried wheat in a planter pot. Only the upper portion of the plant is visible, ensuring focus on the wheat ears. Images were captured along a full circular trajectory around the plant. The plant is positioned against a white background to provide a controlled reconstruction setting. This folder additionally includes instance segmentation annotations in the “Segmentation Mask 1.1” format.

wheat_w_background_w_labels: Dried Wheat (Complex Background, Annotated Images)

This folder contains RGB images of the same dried wheat configuration captured along a full circular trajectory, but positioned against a visually complex background. Instance segmentation masks in “Segmentation Mask 1.1” format are also provided.

This research was supported by the Science Fund of the Republic of Serbia, GRANT No 9029-YF-SAIGE, Project title - LINEARITY.