Data for: Generation of mask from segmented whole plant image dataset of Indian Mustard (Brassica juncea) genotypes for computer vision aided phenotypic trait extraction

The image acquisition was conducted under natural illuminating conditions (9AM to 5PM) at Nanaji Deshmukh Plant Phenomics Centre (NDPPC), ICAR-IARI, New Delhi, during November-December, 2024. High-resolution images of Indian mustard (Brassica juncea) genotypes at 40-45 Days after sowing, at vegetative stage (BBCH16-18) that corresponds to 6 to 8 leaf unfolded stage were captured using a black background for side view and two smartphones one for Top view- Smartphone Camera specification : Samsung Galaxy A14 5G, 50 MP (f/1.8, 26mm wide, 1/2.76", 0.64µm, PDAF) main camera + 2 MP (f/2.4) macro + 2 MP (f/2.4) depth sensor; features include LED flash, panorama, HDR; video recording up to 1080p@30fps mounted on an inhouse built PVC pipe structure and other for Side view Smartphone Camera specification : Samsung Galaxy A05s, 50 MP (f/1.8, wide, autofocus) main camera + 2 MP (f/2.4) macro + 2 MP (f/2.4) depth sensor; features include LED flash, 10x digital zoom; video recording up to 1080p@30/60fps placed on a tripod at a fixed distance. The potted mustard plants were manually rotated to obtain three side view images at 0°, 120° and 240° with a black background. One Top view image was simultaneously captured using topview smartphone. Brassica juncea genotypes were grown in plastic pots filled with 13 kg soil with recommended dose of 100-60-40-20 kg/ha N-P-K-S, respectively. Images of 10 genotypes with two replications and three side views and one top view each were captured in .jpeg format.
This comprehensive dataset facilitate downstream image analysis, binary masks were generated for each image using color-based thresholding techniques. These masks delineate the plant region from the background, providing a clean segmentation layer for further phenotypic trait extraction.The availability of binary masks aligned with multi-view images enables a wide range of analyses within high-throughput phenotyping (HTP) frameworks. Specifically:
• Trait Quantification: The dataset supports automated estimation of morphological features such as leaf area, plant height, rosette diameter, and canopy spread through non-destructive image-based methods.
• 3D Reconstruction Potential: The multi-view nature of the images lends itself to 3D modeling and reconstruction of the plant architecture, enhancing our understanding of growth dynamics over time.
• Digital Image Processing (DIP): The binary segmentation masks simplify tasks such as edge detection, object tracking, and shape analysis, which are foundational in DIP pipelines.
This dataset can serve as a valuable resource for researchers developing computer vision algorithms in plant phenomics, particularly in the context of mustard, and contributes to the broader goals of precision agriculture and crop improvement through AI-assisted tools.