TrashCan 1.0 An Instance-Segmentation Labeled Dataset of Trash Observations

Hong, Jungseok; Fulton, Michael S.; Sattar, Junaed

doi:10.13020/g1gx-y834

Cited by 6 publications

(15 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In contrast to the preceding research that employed a mask R-CNN architecture on the TrashCan dataset, our approach leverages an adapted U-Net model (13,15,18). Despite the variation in architectural choices, a comparative evaluation revealed a notable enhancement in segmentation performance by our model, successfully categorizing all 16 classes, an advancement from an Average Precision of 0.30 attained in the previous work (14).…”

Section: Discussionmentioning

confidence: 91%

“…In this study, we created a deep learning algorithm to detect objects in an image captured by an underwater ROV, with the goal of finding and isolating trash for eventual removal. We utilized a convolutional neural network model developed using the U-Net architecture, which was trained to segment images from the TrashCan dataset (images taken by ROVs) into four distinct classes: trash (with 8 subclasses), animal (with 7 subclasses), plants, and ROV appendages (15). This dataset was composed of 7,212 images of different classes: trash, plants, animals, and ROVs.…”

Section: Resultsmentioning

confidence: 99%

“…This dataset was composed of 7,212 images of different classes: trash, plants, animals, and ROVs. The annotation of the images to create masks was done in a previous research study (15). This process had the potential for human error and biases, which our trained model would eliminate (as it creates masks based off of clearly visible patterns).…”

Section: Resultsmentioning

confidence: 99%

“…The deep learning model in this study was trained and tested using the dataset TrashCan 1.0 (15). Dataset images came from videos sourced from ROVs deployed in the Sea of Japan.…”

Section: Materials and Methods Imaging Datasetmentioning

confidence: 99%

See 3 more Smart Citations

Enhancing marine debris identification with convolutional neural networks

Wahlig,

Gonzales

2024

J Emerg Invest

View full text Add to dashboard Cite

Oceanic debris, predominantly plastic, inflicts catastrophic damage on marine ecosystems, threatens aquatic life through entanglement, strangulation, and starvation, and increasingly accumulates as microplastics within the tissues of marine species consumed by humans. Assessing and mitigating this crisis is complicated by the wide-ranging distribution of waste by ocean currents to remote and deep-sea locations. In this study, we developed a deep learning model to discern and identify components of images captured by an underwater remotely operated vehicle (ROV). We hypothesized that the model would achieve an 80% accuracy. We utilized image segmentation, a process that groups or "masks" all pixels associated with a specific object in the image, to recognize and delineate image components (trash, animal, plant, ROV). We trained our model using the established TrashCan 1.0 dataset, which comprises images captured by ROVs in the Sea of Japan. Our model, a convolutional neural network employing U-Net architecture, formulated feature maps for each object within the images, enabling prediction of object classes. During the testing phase, we compared the model-generated object masks against reference masks to establish the Dice Similarity Coefficient (DSC) for each object class (trash = 0.81 ± 0.38, animal = 0.85 ± 0.35, plant = 0.88 ± 0.31, ROV = 0.86 ± 0.29, overall average = 0.84 ± 0.36). The consistent categorization of image components by our model demonstrates its potential as an effective ROV-borne tool capable of identifying and targeting hitherto inaccessible ocean trash. Our study lays a robust foundation for future research and additional applications.

show abstract

Section: Discussionmentioning

confidence: 91%

Section: Resultsmentioning

confidence: 99%

Section: Resultsmentioning

confidence: 99%

“…The deep learning model in this study was trained and tested using the dataset TrashCan 1.0 (15). Dataset images came from videos sourced from ROVs deployed in the Sea of Japan.…”

Section: Materials and Methods Imaging Datasetmentioning

confidence: 99%

See 2 more Smart Citations

Enhancing marine debris identification with convolutional neural networks

Wahlig,

Gonzales

2024

J Emerg Invest

View full text Add to dashboard Cite

show abstract

“…Four studies (Hegde et al, 2021;Marin et al, 2021;Musić et al, 2020;Wu et al, 2020) retrieved images from internet. While authors of these studies had to manually produce the annotations, one study (Deng et al, 2021) directly utilized the images with annotations from the TrashCan dataset (Hong et al, 2020).…”

Section: Employed Dataset Sourcesmentioning

confidence: 99%