Tobacco Aerial Dataset
Description
To capture aerial dataset of tobacco and weed, a Mavic Mini drone was employed. Between the ages of 15 and 40 days, eight tobacco fields in Mardan, Khyber Pakhtunkhwa, Pakistan, were photographed. The dataset is captured at an average altitude of 4 metres and a ground sampling distance of 0.1 cm per pixel. Pictures were captured for both datasets at a resolution of 1920 × 1080 pixels, however for faster processing, we cropped non-overlapping images to a size of 480 x 352 pixels. Using the MATLAB Image Labeler app, we labelled the photos by hand. A label image in 8-bit unsigned grayscale is produced by the software. The depicted image has the pixel values 0, 1, and 2 assigned to the background, crop, and weed, respectively. Since there are no aerial dataset available publicly to the best of our knowledge, these datasets could be utilized to support further research for Tobacco crop. Tobacco dataset is captured in eight different drone fly campaigns on eight different fields as shown in table below. Campaign No. Number of images Timing Around/ Date Captured Soil Condition 1 864 2:30pm/09April 21 After irrigation 2 936 3:30pm/07April 21 Before irrigation 3 120 3:38pm/07April 21 After irrigation 4 120 5:52pm/07April 21 Before irrigation 5 120 6:33pm/07April 21 After irrigation 6 120 3:27pm/09April 21 After irrigation 7 120 3:43pm/09April 21 Before irrigation 8 120 3:59pm/09April 21 After irrigation Citation Request: if you use these datasets in your research or projects by any means, please cite following publications. 1) Patch-wise weeds coarse segmentation mask from aerial imagery of sesame crop (Published in Computers and Electronics in Agriculture 2022, HEC Recognized W category, Impact factor 6.757, Q1) 2) Towards automated weed detection through two-stage semantic segmentation of tobacco and weed pixels in aerial Imagery (Published in Smart Agricultural Technology (A companion journal of Computers and Electronics in Agriculture)) 3) A Patch-Image Based Classification Approach for Detection of Weeds in Sugar Beet Crop (Published in IEEE Access, Impact factor 3.1, Q1) Acknowledgement Request This work is funded by the Higher Education Commission of Pakistan and the National center for Robotics and Automation (DF-1009–31). Please Acknowledge. Find More data and research papers in related links attached.
Files
Steps to reproduce
Tobacco Aerial Datasets is the main folder and it has eight subfolders named, Campaign no.1-8 (one to eight). All these folders contains self-descriptive folders named Detected vegetation (contains 1080P images where background is removed), Patch images (Contains cropped 480*352 size images and corresponding labels), Binary vegetation mask (contains binary image which show background and vegetation), Original 1080P images (contains original images) and Labelled 1080P images (Contains labels of original 1080P images). Patch images folder contains three subfolders data (Cropped original 480*352 patch images), mask(corresponding 480*352 patch labels) and maskref (480*352 size labels just for visualization purpose). Contact Dr. Syed Imran Moazzam Shah in case of confusion drimoazz@gmail.com imoazzam@ceme.nust.edu.pk +92-313-7233382