PlastOPol dataset aiming of giving to the computer science and environmental science communities a new set of images with the presence of litter in several types of environments. We hope that PlastOPol serves as a basis for the proposal of automatic detection methods which can support the furthering Sensors 2022, 1, 0 6 of 20 of research on litter in the environment. The images were collected by the Marine Debris Tracker available under an open access Creative Commons Attribution license. Building the dataset involved the meticulous task of labeling each litter instance in each image. PlastOPol is a one-class labeled dataset, where all the data corresponds to the “litter” class as its super-category. This dataset has 2418 images collected by the Marine Debris Tracker with a total of 5300 instances of litter. Each instance is wrapped within a rectangular bounding box represented by four values (x1, y1, width, and height), where (x1, y1) corresponds to the upper left corner of the bounding box. If you use this dataset, please cite our paper: @Article{Cordova2022Sensors, AUTHOR = {Córdova, Manuel and Pinto, Allan and Hellevik, Christina Carrozzo and Alaliyat, Saleh Abdel-Afou and Hameed, Ibrahim A. and Pedrini, Helio and Torres, Ricardo da S.}, TITLE = {Litter Detection with Deep Learning: A Comparative Study}, JOURNAL = {Sensors}, VOLUME = {22}, YEAR = {2022}, NUMBER = {2}, ARTICLE-NUMBER = {548}, URL = {https://www.mdpi.com/1424-8220/22/2/548}, ISSN = {1424-8220}, DOI = {10.3390/s22020548} }

PlastOPol dataset aiming of giving to the computer science and environmental science communities a new set of images with the presence of litter in several types of environments. We hope that PlastOPol serves as a basis for the proposal of automatic detection methods which can support the furthering Sensors 2022, 1, 0 6 of 20 of research on litter in the environment. The images were collected by the Marine Debris Tracker available under an open access Creative Commons Attribution license. Building the dataset involved the meticulous task of labeling each litter instance in each image. PlastOPol is a one-class labeled dataset, where all the data corresponds to the “litter” class as its super-category. This dataset has 2418 images collected by the Marine Debris Tracker with a total of 5300 instances of litter. Each instance is wrapped within a rectangular bounding box represented by four values (x1, y1, width, and height), where (x1, y1) corresponds to the upper left corner of the bounding box.