Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 1
Encontrada(s) 1 página(s)

DIGITAL.CSIC. Repositorio Institucional del CSIC
Datos de investigación Set de datos (Dataset)

High- and Low-Quality RGB Image Dataset for Weed Species Identification in Vineyards

  • Emmi, Luis Alfredo
  • Fernández Saavedra, Roemi
[Description of methods used for collection/generation of data] The data were acquired using the RGB camera TRI032S-CC RGB from Lucid Vision Labs equipped with the SV- 04514V lens (resolution: 5 MP; FoV: 59.4° × 79°). Two cameras were used, one on the right and one on the left of a mobile robotic platform, which navigated autonomously thro2., [Methods for processing the data] - Image Acquisition: RGB images were captured (left and right cameras) in a vineyard environment. Each image was stored in both high-quality (HQ) and low-quality (LQ) formats. - Image Preprocessing: Images were either used in full resolution or cropped to focus on specific regions of interest. Additionally, ORB (Oriented FAST and Rotated BRIEF) feature detection was applied to extract keypoints for targeted analysis. - Species Identification: Each image was submitted to the Pl@ntNet API for automated plant species identification. The API returns a ranked list of candidate species with associated confidence scores. The raw responses were stored in .pkl files, separately for left and right cameras. - Data Structuring: The Pl@ntNet responses were organized into Python dictionaries, indexed by image filename, and saved in structured folders based on project, image quality, and analysis method. Certain images were subjected to blurring in specific regions to anonymize individuals and obscure sensitive information, including vehicle license plates, in compliance with data protection and ethical guidelines.ugh a vineyard, obtaining images of the intra-row space., This dataset contains RGB images and corresponding species identification results aimed at supporting weed detection in vineyard environments. The images were captured using a dual-camera system (left and right) mounted on an autonomous mobile robot. Each camera recorded both high-quality (HQ) and low-quality (LQ) images, which were later analyzed using the Pl@ntNet API. The dataset is organized into two main components: an Images folder with raw RGB captures, and a Data folder containing structured analysis results in Python pickle format. These results include full image analysis, cropped image analysis, and ORB (Oriented FAST and Rotated BRIEF) feature-based analysis. The data were collected on July 4, 2024, in a vineyard located at 41°59'09.9"N, 8°46'29.9"W, as part of the AGROSUS project funded by the European Union’s Horizon research and innovation programme (Grant No. 101084084)., This dataset is part of a the AGROSUS project that has received funding from the European Union’s Horizon research and innovation programme under grant agreement No 101084084., This dataset is organized into two main directories: Images and Data, each containing structured subfolders that reflect the experimental setup and data processing pipeline. - Images Folder The Images directory contains RGB images captured by a color camera system composed of a left and a right camera located in a mobile autonomous robot. It is structured as follows: Left_Camera_RGB/ HQ/: High-quality images from the left camera. LQ/: Low-quality images from the left camera. Right_Camera_RGB/ HQ/: High-quality images from the right camera. LQ/: Low-quality images from the right camera. These images were used as input for species identification using the Pl@ntNet API. - Data Folder The Data directory contains the results of the image analysis and species identification process. It is organized by project and image quality: Weeds_HQ/ Weeds_LQ/ SouthW_HQ/ SouthW_LQ/ Each HQ and LQ folder contains three subdirectories: Full_Images_Results/: Raw results from full image analysis. Full_Cropped_Images_Results/: Results from cropped image analysis. ORB_Results/: Results using ORB (Oriented FAST and Rotated BRIEF) feature-based analysis. Within each of these subdirectories, there are two .pkl files: --camera_left.pkl --camera_right.pkl These files are Python dictionaries where each key corresponds to an image filename, and the value is the raw response from the Pl@ntNet API, containing species identification data., Peer reviewed