Supervised deep learning methods have been widely employed to detect floating macroplastic litter (>5 mm) in (fresh)water bodies. However, few studies used them to quantify floating litter fluxes in rivers with wide cross-sections, that is important for pollution assessment. Additionally, commonly used supervised learning (SL) models rely on extensive labeled data, that is time-consuming and expensive to obtain. Moreover, regardless of the model type, current deep learning models for litter detection usually fail to correctly identify small litter items. To overcome these issues, we propose a semi-supervised learning (SSL)-based framework combined with Slicing Aided Hyper Inference (SAHI) for quantifying cross-sectional floating litter fluxes in rivers. The framework includes four steps: (a) collecting camera images of river surfaces from multiple locations across the river, (b) developing a robust litter detection model using SSL, (c) applying this model with SAHI to detect litter items in images, and (d) post-processing the detection results to quantify fluxes. The SSL method involves: (i) self-supervised pre-training of a ResNet50 on a large amount of unlabeled data, and (ii) supervised fine-tuning of a Faster R-CNN with the ResNet50 backbone on a limited amount of labeled data. We evaluated the in-domain detection performance of SSL models with varying pre-training epochs and pre-training dataset sizes, using images from waterways of The Netherlands, Indonesia and Vietnam, that were used for model pre-training and fine-tuning. Additionally, we assessed the zero-shot out-of-domain detection performance of SSL models and litter flux quantification performance of the proposed framework on a Vietnam case study, that was not used for model development. We benchmarked our results against the SL methods and human visual counting. The results show that SSL models benefit from longer pre-training time and larger pre-training dataset, achieving an in-domain F1-score increase of 0.2 and a zero-shot out-of-domain increase of up to 0.14, over baseline SL benchmarks. Furthermore, the SAHI method correctly identifies 45 additional small litter items (areas < 1,000 cm2), improving the F1-score by up to 0.19, compared to the results obtained without SAHI. The flux measurement results indicate that the SSL-based framework substantially underestimates fluxes by a factor of 3–4 compared to human measurements, due to missed detections of transparent litter items and items entrapped in water hyacinths. However, it estimates nearly twice the fluxes of the baseline SL-based framework, aligning more closely with human measurements. These findings highlight the potential of SSL-based framework to enhance litter flux measurement. Scaling it with broader datasets could significantly advance global-scale litter monitoring systems.

Researchers and practitioners have extensively utilized supervised Deep Learning methods to quantify floating litter in rivers and canals. These methods require the availability of large amount of labeled data for training. The labeling work is expensive and laborious, resulting in small open datasets available in the field compared to the comprehensive datasets for computer vision, e.g., ImageNet. Fine-tuning models pre-trained on these larger datasets helps improve litter detection performances and reduces data requirements. Yet, the effectiveness of using features learned from generic datasets is limited in large-scale monitoring, where automated detection must adapt across different locations, environmental conditions, and sensor settings. To address this issue, we propose a two-stage semi-supervised learning method to detect floating litter based on the Swapping Assignments between multiple Views of the same image (SwAV). SwAV is a self-supervised learning approach that learns the underlying feature representation from unlabeled data. In the first stage, we used SwAV to pre-train a ResNet50 backbone architecture on about 100k unlabeled images. In the second stage, we added new layers to the pre-trained ResNet50 to create a Faster R-CNN architecture, and fine-tuned it with a limited number of labeled images (≈1.8k images with 2.6k annotated litter items). We developed and validated our semi-supervised floating litter detection methodology for images collected in canals and waterways of Delft (the Netherlands) and Jakarta (Indonesia). We tested for out-of-domain generalization performances in a zero-shot fashion using additional data from Ho Chi Minh City (Vietnam), Amsterdam and Groningen (the Netherlands). We benchmarked our results against the same Faster R-CNN architecture trained via supervised learning alone by fine-tuning ImageNet pre-trained weights. The findings indicate that the semi-supervised learning method matches or surpasses the supervised learning benchmark when tested on new images from the same training locations. We measured better performances when little data (≈200 images with about 300 annotated litter items) is available for fine-tuning and with respect to reducing false positive predictions. More importantly, the proposed approach demonstrates clear superiority for generalization on the unseen locations, with improvements in average precision of up to 12.7%. We attribute this superior performance to the more effective high-level feature extraction from SwAV pre-training from relevant unlabeled images. Our findings highlight a promising direction to leverage semi-supervised learning for developing foundational models, which have revolutionized artificial intelligence applications in most fields. By scaling our proposed approach with more data and compute, we can make significant strides in monitoring to address the global challenge of litter pollution in water bodies.

Supervised Deep Learning (DL) methods have shown promise in monitoring the floating litter in rivers and urban canals but further advancements are hard to obtain due to the limited availability of relevant labeled data. To address this challenge, researchers often utilize techniques such as transfer learning (TL) and data augmentation (DA). However, there is no study currently reporting a rigorous evaluation of the effectiveness of these approaches for floating litter detection and their effects on the models' generalization capability. To overcome the problem of limited data availability, this work introduces the “TU Delft—Green Village” dataset, a novel labeled dataset of 9,473 camera and phone images of floating macroplastic litter and other litter items, captured using experiments in a drainage canal of TU Delft. We use the new dataset to conduct a thorough evaluation of the detection performance of five DL architectures for multi-class image classification. We focus the analysis on a systematic evaluation of the benefits of TL and DA on model performances. Moreover, we evaluate the generalization capability of these models for unseen litter items and new device settings, such as increasing the cameras' height and tilting them to 45°. The results obtained show that, for the specific problem of floating litter detection, fine-tuning all layers is more effective than the common approach of fine-tuning the classifier alone. Among the tested DA techniques, we find that simple image flipping boosts model accuracy the most, while other methods have little impact on the performance. The SqueezeNet and DenseNet121 architectures perform the best, achieving an overall accuracy of 89.6 and 91.7%, respectively. We also observe that both models retain good generalization capability which drops significantly only for the most complex scenario tested, but the overall accuracy raises significantly to around 75% when adding a limited amount of images to training data, combined with flipping augmentation. The detailed analyses conducted here and the released open source dataset offer valuable insights and serve as a precious resource for future research.

Plastic pollution in water bodies is an unresolved environmental issue that damages all aquatic environments, and causes economic and health problems. Accurate detection of macroplastic litter (plastic items >5 mm) in water is essential to estimate the quantities, compositions and sources, identify emerging trends, and design preventive measures or mitigation strategies. In recent years, researchers have demonstrated the potential of computer vision (CV) techniques based on deep learning (DL) for automated detection of macroplastic litter in water bodies. However, a systematic review to describe the state-of-the-art of the field is lacking. Here we provide such a review, and we highlight current knowledge gaps and suggest promising future research directions. The review compares 34 papers with respect to their application and modeling related criteria. The results show that the researchers have employed a variety of DL architectures implementing different CV techniques to detect macroplastic litter in various aquatic environments. However, key knowledge gaps must be addressed to overcome the lack of: (i) DL-based macroplastic litter detection models with sufficient generalization capability, (ii) DL-based quantification of macroplastic (mass) fluxes and hotspots and (iii) scalable macroplastic litter monitoring strategies based on robust DL-based quantification. We advocate for the exploration of data-centric artificial intelligence approaches and semi-supervised learning to develop models with improved generalization capabilities. These models can boost the development of new methods for the quantification of macroplastic (mass) fluxes and hotspots, and allow for structural monitoring strategies that leverage robust DL-based quantification. While the identified gaps concern all bodies of water, we recommend increased efforts with respect to riverine ecosystems, considering their major role in transport and storage of litter.