An intelligent way for discerning plastics at the shorelines and the seas

Abstract

Irrespective of how plastics litter the coastline or enter the sea, they pose a major threat to birds and marine life alike. In this study an Artificial Intelligence tool was used to create an image classifier based on a Convolutional Neural Network architecture that utilises the Bottleneck Method. The trained Bottleneck Method image classifier was able to categorise plastics encountered either at the shoreline or floating at the sea surface into eight (8) distinct classes, namely, plastic bags, bottles, buckets, straws, derelict nets, fish and other objects. Discerning objects with a success rate of 90%, the proposed Deep Learning approach constitutes a leap towards the automatic identification of plastics at the coastline and the sea. Training and testing loss and accuracy results for a range of epochs and batch sizes has lent credibility to the proposed method. Results originating from a resolution sensitivity analysis demonstrated that the prediction technique retains its ability to correctly identify plastics even when image resolution was downsized by 75%. Intelligent tools, such as the one suggested here can replace manual sorting of macroplastics from human operators revealing, for the first time, the true scale of the amount of plastic polluting our beaches and the seas.

A Citizen Science Unmanned Aerial System Data Acquisition Protocol and Deep Learning Techniques for the Automatic Detection and Mapping of Marine Litter Concentrations in the Coastal Zone

“…In contrast with most previous publications [22,34,[46][47][48], this study used a significantly larger training and validation dataset acquired from five different beach environments with complex background characteristics and litter concentrations. Additionally, the evaluation of the deep learning models' generalization ability in a completely new beach environment expands the geographical transferability of our approach to new and unknown beaches.…”

“…In this study, we utilized the valuable knowledge acquired by several CNNs on the classification task of the ImageNet Large-Scale Visual Recognition Challenge (ILSVRC) [63] through the transfer learning approach to re-purpose them towards ML recognition from UAS aerial images. Regarding their performance on the ILSVRC classification task, their pioneering improvements, and their use in ML detection [22,47,48], we selected 5 CNN architectures from previous studies: (a) 2 plain architectures [49] (VGG-16 and VGG-19) and (b) 3 densely connected variations [64] (DenseNet-121, DenseNet-169, DenseNet-201).…”

“…Automatic detection and reporting are essential for such a procedure; thus, several attempts are using artificial intelligence (AI) approaches for fast and accurate responses. Recent works have explored the viability of an UAS-based approach to detect, identify, and classify marine litter in coastal areas by applying automated detection using machine learning methods or deep learning techniques [22,34,[45][46][47][48].…”

“…The test dataset consisted of 165 images and achieved an overall accuracy of 86%. Kylili et al [47] attempted to improve their previous approach by extending the number of classes from three to eight, which increased the training and validation image-sets. In order to evaluate CNN performance, the authors used a testing image-set of 400 samples, giving an overall accuracy of 90%, which improved their previous result by 4%.…”

Smart Fishery: A Systematic Review and Research Agenda for Sustainable Fisheries in the Age of AI

“…Sustainable fishery is also related to environmental monitoring, which is reflected in the overlap of papers (see Table 4). Kylili et al [38] (p. 42632) locate the scope of smart fishery in "epidemic of plastics entering the sea warrants urgent action if humanity is to stave off a collapse in fish stocks". They demonstrate that pollution is an urgent and necessary consideration regarding fish abundance.…”