Potato Purple Top (PPT) disease dataset for training classification Convolutional Neural Networks (CNN) such as ResNet and EfficientNet (090626)

AbstractThe purpose of this project is to evaluate the effectiveness of two convolutional neural networks (ResNet and EfficientNet) in identifying symptoms of Potato Purple Top disease (PPT) by analyzing frames obtained from video recordings made from an drone (UAV). As part of the results of the aforementioned project, a dataset comprising a total of 13.884 frames is presented, which can be used as input for training convolutional neural networks for classification in Potato Purple Top (PPT). Dataset Information The dataset under consideration contains a total of 13.884 frames, extracted from 3 hours, 20 minutes, and 37 seconds of video footage captured by a DJI AIR S2 drone camera. The camera's resolution is specified as 5K Ultra HD at 30 frames per second (1920 x 1080), and the footage was captured using an automated process implemented in OpenCV. Initially, the native frame rate of each video was obtained, from which a sampling interval was calculated in order to uniformly extract three frames per second. The selection was made sequentially, with every Nth frame being retained according to the defined interval. The resulting files were stored using a nomenclature based on the video timestamp and an ordinal identifier for the frame within each second. All videos were captured from the UAV, with a distance ranging from 3 to 5 metres above 11 potato crops where PPT had previously been identified during the period from March 2024 to September 2025.The visual categorization of potato plants with PMP was performed using digital images taken in the field under natural lighting conditions. The plants were classified into two categories: (i) Healthy and (ii) Diseased, based on previously defined visual criteria. The dataset is divided into Training, Validation, and Test classes in an 80:10:10 ratio, as follows: Class Healthy Disease TOTAL Training 1.455 8.397 9.852 ValIdation 1.001 1.028 2.029 Test 1.001 1.002 2.003 TOTAL 3.457 10.427 13.884