Volume 5 - 2023 | <a class="ArticleLayoutHeader__info__doi" href="https://doi.org/10.3389/frwa.2023.1298465">https://doi.org/10.3389/frwa.2023.1298465</a>
This article is part of the Research TopicApplied Machine Learning for River Components Studies<a class="Link Link--linkType Link--maincolor Link--medium Link--icon Link--chevronRight Link--right" aria-label="View all 3 articles" href="https://www.frontiersin.org/research-topics/54484/applied-machine-learning-for-river-components-studies/articles" target="_self" data-event="customLink-linkType-a_viewAll3Articles">View all 3 articles</a>
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
researchers often utilize techniques such as transfer learning (TL) and data augmentation (DA)
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&#x00027; generalization capability
To overcome the problem of limited data availability
this work introduces the &#x0201C;TU Delft&#x02014;Green Village&#x0201D; 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
we evaluate the generalization capability of these models for unseen litter items and new device settings
such as increasing the cameras&#x00027; height and tilting them to 45&#x000B0;
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
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%
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
The detailed analyses conducted here and the released open source dataset offer valuable insights and serve as a precious resource for future research
While reported results are encouraging, DL-based detection of floating litter faces several challenges, as recently highlighted by <a href="#B9">Gnann et al. (2022)</a> and <a href="#B16">Jia et al. (2023)</a>
One major challenge is the need of large annotated datasets to train and validate robust DL models
Acquiring a sufficiently large dataset can be time-consuming
training models from scratch or with non-augmented datasets)
did not generalize well to a different location of the same city
resulting in a drop in precision from 68.7 to 54.0%
This degradation in performance was attributed to the presence of a large amount of organic material (e.g.
which was not accounted for during training
no study exists on DL-based litter detection reporting a rigorous comparison of TL strategies
DA techniques and their effects on generalization
we contribute to closing this gap and reduce the issue of data availability by performing a thorough analysis on a novel dataset of floating litter
The key contributions of this paper can be summarized as follows:
We release the &#x0201C;TU Delft&#x02014;Green Village&#x0201D; (TUD-GV) dataset
a novel labeled dataset of camera and phone images of floating litter collected from semi-controlled experiments in a drainage canal of the TU Delft Campus
We train and test five deep learning architectures for multi-class image classification (ResNet50
and evaluate the benefits of multiple TL strategies and DA techniques on detection performance;
We perform tailored experiments to investigate and improve the out-of-sample generalization capability of the best performing architectures
considering unseen litter items and new device settings (i.e.
increasing the cameras&#x00027; height and tilting the viewing angle to 45&#x000B0;)
This study is aiming primarily at researchers developing novel DL methods for improved detection and quantification of floating litter in rivers and canals. The insights summarized above can contribute to other stakeholders such as various technology providers, consultants, governmental organizations, and Non-Governmental Organizations (NGOs) to develop more robust DL models for cleaning up campaigns and mitigation of environmental pollution (<a href="#B16">Jia et al., 2023</a>)
The remainder of the paper is structured as follows
Section 3 presents the methodology used in this study
and proposed approaches to improve generalization capability
Section 4 describes three sets of experiments
we present and discuss the experimental results
Monitoring setup at The Green Village: (A) view from the top with the four different filming locations (1&#x02013;4) on the bridge; (B) details of some camera installation on Locations 1 and 2
Examples of images from the TUD-GV dataset captured from four device setups including (A) 2.7 m/0&#x000B0;
The captions for the four images are (A) no litter
The image (C) was cropped to omit the bridge
The TUD-GV dataset described above is one of the most complete and accessible datasets for detecting floating litter using computer vision, and a much needed supplement to existing datasets made available by other researchers (<a href="#B43">van Lieshout et al., 2020</a>; <a href="#B46">Wolf et al., 2020</a>)
The reader is referred to the literature for more details on the employed architectures
A typical CNN for image classification consists of several convolutional blocks and a classifier. The convolutional blocks are made up of convolutional and pooling layers, which are used to extract features from images. The classifier typically consists of fully connected dense layers that are used to classify images based on the features extracted by the convolutional base (<a href="#B33">Subramanian et al., 2022</a>)
we replaced the original classifier in each CNN architecture with a global average pooling layer followed by a dense layer with a softmax activation function for multi-class classification (i.e.
Global average pooling summarizes the feature maps produced by the convolutional base to reduce overfitting and computational costs
We evaluated the benefits of the most common transfer learning strategies (<a href="#B10">Guo et al., 2020</a>): (1) fine-tuning the classifier alone (FTC), and (2) fine-tuning all layers (FTAL). We evaluated the effect of transferring features learned on the ImageNet IC task to floating litter detection, a common approach in the field (<a href="#B16">Jia et al., 2023</a>)
The ImageNet dataset is a widely used benchmark dataset for IC tasks
balloon and strawberry) and over 14 million images
we first loaded the model pre-trained on ImageNet
then replaced and fine-tuned the classifier on the TUD-GV dataset while freezing the convolutional base (i.e.
we fine-tuned all layers of the model on the TUD-GV dataset after loading the ImageNet weights as the starting point and replacing the classifier
We compared the effectiveness of FTC and FTAL against the performances obtained by training the models from scratch
Each DA method mentioned above was applied to generate three new images for each original training image
MIX DA includes all images generated by the other four DA methods
resulting in a total of 12 new images for each original training image
Examples of data augmentation techniques used
(Left) An original image; (Right) images generated by performing horizontal flipping (top row
combined horizontal and vertical flipping (middle row
and adding salt and pepper noise (bottom row
we tested the generalization capability of the trained models to different types of litter from various device setups (camera height/angle)
We trained the models on images of canal litter from the 2.7 m/0&#x000B0; setup
and evaluated their out-of-sample performances on images of household waste from 2.7 m/0&#x000B0;
we applied three methods to improve the models&#x00027; generalization capability
which are (1) using the best DA method emerging from our study on data augmentation; (2) adding new images to the training dataset (ANI); and (3) using the best DA method after adding new images to the training dataset (ANI-DA)
All the new images added to the original training dataset still featured canal litter
and 4 m/45&#x000B0; setups to better represent the out-of-sample distributions
In this study, we conducted three experiments to: (1) compare the performances of the five DL architectures, with and without transfer learning; (2) assess the performance boost of the five different DA approaches on the two best performing models from (1); and (3) evaluate and improve the generalization capability of the best models for unseen litter (e.g., household waste) and different device setups. <a href="#F4">Figure 4</a> shows the flowchart of three experiments
mixing all the four aforementioned techniques; DA
adding new images to original training dataset; ANI-DA
With the first experiment we compared the detection performances of the five chosen DL architectures (i.e., ResNet, InceptionV3, DenseNet121, MobileNetV2, and SqueezeNet) and we assessed the benefits of the FTC and FTAL strategies described in Section 3.2. We used a shuffled subset of 4,005 images with canal litter for model development, subdivided into training, validation, and test datasets following the 80/10/10 split detailed in <a href="#T2">Table 2</a>
Ratios between the different classes is kept constant across the different datasets
All images have been recorded from the action cameras with the 2.7 m/0&#x000B0; setup
We applied the five different DA techniques in Section 3.3 to the two top performing baseline models emerging from Experiment 1. These were retrained on the augmented datasets yielded by applying each DA techniques, resulting in 12,812 or 41,639 training images (MIX DA). For a fair comparison against the baselines, we used the same Validation and Test datasets of Experiment 1 (see <a href="#T2">Table 2</a>)
We created the TrainANI dataset by adding 1,523 images (4,726 total) of canal litter to the Train dataset of Experiment 1
from the three missing device setups (2.7 m/45&#x000B0;
The TrainANI &#x02212; DA dataset was created by performing DA on TrainANI
resulting in a total of 18,904 training images
We validated the models for both ANI and ANI-DA cases on the ValidationANI dataset
obtained by adding 188 images of canal litter to the Validation dataset of Experiment 1
We compared these models against the baselines of Experiment 1 and the best performing models with DA of Experiment 2
The weight of each class was calculated as the ratio of the total number of images to the number of images in that particular class
we repeated the training 10 times for each model in all experiments
All the results reported in Section 5 are mean values calculated from these runs
unless we discuss the outcomes of misclassification analysis
which we conducted on the best performing models out of the 10 runs
We implemented the DL architectures using the Python programming language (version 3.8.5) and the Keras DL framework (version 2.6.0). We used the implementations and pre-trained weights from tf.keras.applications for all architectures, except for SqueezeNet.<a id="note1a"></a><a class="footnoteanchor" href="#note1">1</a> Model development was performed on a local NVIDIA GeForce RTX 3090 GPU (24GB)
To evaluate model performances of floating litter detection, we used four metrics commonly employed in multi-class IC tasks: overall accuracy (OA), precision, recall, and F1-score (<a href="#B16">Jia et al., 2023</a>)
We used OA to summarize model performance across all classes
This metric measures the percentage of correctly identified images out of the total images in the dataset
where N is the total number of images; K represents the number of classes; and Ci
i denotes the number of images that are actually in class i and identified as such
recall and F1-score to assess the performances for each class
Precision for class i is written as follows:
where TPi (True Positive) represents the number of correctly classified images of class i; and FPi (False Positive) represents the number of images misclassified as class i
Recall for class i is expressed as follows:
where FNi (False Negative) represents the number of images that are actually in class i but classified as other classes
Precision reflects how accurate is the model in identifying relevant samples
It identifies the percentage of correctly identified positive samples over the total identified positive samples
recall represents the model&#x00027;s ability to identify all relevant samples
It is the percentage of correctly identified positive samples over the total positive samples
F1-score combines the two metrics by computing their harmonic mean
<a href="#T4">Table 4</a> reports the average training time and OA on the Test dataset for the five architectures trained from scratch or fine-tuned after transfer learning. In this table, we only reported the learning rate that yields the best average OA on the validation set for each architecture. The full evaluation of the five architectures with all tested learning rates can be found in <a href="#SM1">Supplementary material</a>
The FTAL method consistently outperforms the other methods regardless of the architecture
we obtained OA ranging from 85.0 to 87.6% on the Test set
Training models from scratch performs slightly worse than the FTAL method
with OAs varying between 62.3 and 73.3% depending on the architecture
switching from FTC to FTAL with ResNet50 yields a significant improvement of &#x0002B;22.7% in OA
the FTC method consistently takes the least training time
costing between 2 and 9 s for each training epoch
That is &#x0007E;2 to 5 times faster than using the FTAL method or training the models from scratch
This is expected since training or fine-tuning the entire network takes significantly more time than fine-tuning the classifier alone
and overall accuracy of all architectures for Experiment 1
Our findings are similar to those reported by <a href="#B24">Marin et al. (2021)</a> for a study on CNN architectures detecting underwater litter
The authors classified images into six classes: glass
the FTAL strategy proved more successful than resorting to FTC
with best performance on the test dataset of OA = 91.4% compared to 83.0%
We observed a relatively high number of errors for images belonging to the &#x0201C;moderate litter&#x0201D; class
are sometimes confused with &#x0201C;little litter&#x0201D; (14 case) or &#x0201C;lots of litter&#x0201D; (5 case)
resulting in the lowest F1-scores for this class across all architectures
F1-score per class of all architectures trained using the FTAL strategy for Experiment 1
Confusion matrix of the best performing DenseNet121 trained with the FTAL strategy for Experiment 1
Performances of SqueezeNet and DenseNet121 using different DA techniques for Experiment 2
The horizontal dashed line represents the OA of the baseline models
and (5) mixing the four above-mentioned techniques (MIX DA)
This trade-off should be considered when choosing a model for a particular specific litter detection task
Although the generalization capability in this case is already satisfactory
DenseNet121 models trained with DA and ANI-DA show significant increases is OA of &#x0002B;5.4 and &#x0002B;6.2%
while ANI alone does not provide a similar boost
Lesser improvements are also measured for SqueezeNet
it requires the time-consuming and costly collection of new data
simple flipping augmentation may be the most cost-effective method to improve the generalization capability under the same device setup
Out-of-sample generalization performances of SqueezeNet (A) and DenseNet121 (B) on Experiment 3
featuring baseline models and models leveraging techniques for improved generalization
Comparison performed on datasets of four device setups (2.7 m/0&#x000B0;
and 4 m/45&#x000B0;) with household litter
The SqueezeNet and DenseNet121 baselines exhibit good performances on Test2.7m/45&#x000B0;, with OA of 90.7 and 83.8%, respectively. Overall, results are similar or better than for the simpler Test2.7m/0&#x000B0; because Test2.7m/45&#x000B0; consists exclusively of images taken in cloudy weather. Sunny weather images are harder to classify due to the presence of sun glints (<a href="#B15">Jakovljevic et al., 2020</a>)
The three approaches significantly improve the generalization capability of DenseNet121
we could not improve the performances of SqueezeNet further
suggesting that this small architecture cannot incorporate larger amount of data effectively
SqueezeNet still retains good generalization capability with OAs above 86.7%
Common misclassified examples in the Test4m/0&#x000B0; and the Test4m/45&#x000B0; datasets for the best baseline SqueezeNet model
Common misclassification include identifying sun glints as litter
and detection of background objects or external items
The ANI method outperforms simple flipping augmentation on these harder datasets
with improvements of around &#x0002B;11% for both architecture in each setup
While flipping grants significant increases of up to &#x0002B;9.4% in Test4m/0&#x000B0;
it fails to support generalization for the more complex Test4m/45&#x000B0;
This suggests that simple data augmentation fails to boost generalization when the out-of-sample distribution is significantly different from the training one (e.g.
collecting new data from the new setup is necessary to achieve satisfactory performances
Performing DA after gathering new images can result in further improvements
as demonstrated in Test4m/0&#x000B0; for both SqueezeNet and DenseNet121 (i.e.
respectively) and in Test4m/45&#x000B0; for DenseNet121 (OA = 77.7%)
providing more meaningful context than ImageNet or other general purpose datasets
the development of these models can benefit directly from the insights provided by this study
Supervised Deep Learning methods are increasingly used for automatic detection of floating litter in aquatic environments
These methods require large amounts of carefully labeled data to reach satisfactory performances and exhibit strong generalization capability
there is a major lack of available datasets for researchers and practitioners willing to develop and test their models
we introduced the &#x0201C;TU Delft-Green Village&#x0201D; (TUD-GV) dataset
a collection of almost 10,000 labeled images of different floating litter recorded from semi-controlled experiments on a small drainage canal in the TU Delft Campus
We demonstrated the usefulness of our data by carrying out a thorough comparison of different DL architectures
and approaches to improve out-of-sample generalization for unseen litter and device setups
The main findings of our work are as follows:
(1) DenseNet121 and SqueezeNet proves the most successful architectures thanks to their innovative designs that allow for efficient feature discovery for floating litter classification with limited trainable parameters
(2) We obtained the highest accuracy by loading models pre-trained on ImageNet
and fine-tuning the entire network on floating litter images
The benefits of this approach in terms of detection performance outweigh the shorter training times required by fine-tuning the classifier alone
Transferring the convolutional base from ImageNet seems a better approach than training the models from scratch
(3) We recommend flipping data augmentation (DA) to improve model detection performances at relatively low cost
since the additional images are easy to generate
and maintain high fidelity to the original labels while providing extra training information
and adding noise do not show a significant improvement in detecting floating litter
(4) The trained models generalize well to similar conditions
such as detecting unseen litter items from images captured at the same height
Flipping data augmentation may boost generalization performance in these circumstances
but it is insufficient when transferring to more complex scenarios (e.g.
different camera heights and different viewing angle)
We demonstrated that adding a limited amount of images from these new settings to the original training dataset can substantially improve generalization in these cases
scaling and perspective skewing transformations may improve generalization for the most complex scenario featured in TUD-GV
This investigation should be paired with an in-depth analysis to evaluate the variation of litter detection performance with respect to item size and camera inclination
Multi-scale data augmentation can also improve generalization performances for applications using cameras with different resolutions
The code for this study is available on <a href="https://github.com/TianlongJia/deep_plastic">https://github.com/TianlongJia/deep_plastic</a>. The &#x0201C;TU Delft-Green Village&#x0201D; dataset is available for download from Zenodo at <a href="https://doi.org/10.5281/zenodo.7636124">https://doi.org/10.5281/zenodo.7636124</a>
The author(s) declare financial support was received for the research
This work was supported by China Scholarship Council (No
202006160032) and the Directorate-General for Public Works and Water Management of The Netherlands (Rijkswaterstaat)
AV and RV were employed by Noria Sustainable Innovators
The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations
Any product that may be evaluated in this article
or claim that may be made by its manufacturer
is not guaranteed or endorsed by the publisher
The Supplementary Material for this article can be found online at: <a href="https://www.frontiersin.org/articles/10.3389/frwa.2023.1298465/full#supplementary-material">https://www.frontiersin.org/articles/10.3389/frwa.2023.1298465/full#supplementary-material</a>
1. <a class="footnotetextanchor" href="#note1a" title="">^</a><a style="color:grey;" href="https://github.com/rcmalli/keras-squeezenet">https://github.com/rcmalli/keras-squeezenet</a>
<a href="https://doi.org/10.30630/joiv.2.4.151" target="_blank">CrossRef Full Text</a> | <a href="http://scholar.google.com/scholar_lookup?author=J.+Azzeh&#x00026;author=B.+Zahran&#x00026;author=Z.+Alqadi+&#x00026;publication_year=2018&#x00026;title=Salt+and+pepper+noise%3A+effects+and+removal&#x00026;journal=Int.+J.+Inf.+Visualiz&#x00026;volume=2&#x00026;pages=252-256" target="_blank">Google Scholar</a>
Predicted growth in plastic waste exceeds efforts to mitigate plastic pollution
An embeddable algorithm for automatic garbage detection based on complex marine environment
<a href="http://scholar.google.com/scholar_lookup?author=J.+Deng&#x00026;author=W.+Dong&#x00026;author=R.+Socher&#x00026;author=L.+-J.+Li&#x00026;author=K.+Li&#x00026;author=L.+Fei-Fei+&#x00026;publication_year=2009&#x00026;title=&#x0201C;Imagenet%3A+a+large-scale+hierarchical+image+database,&#x0201D;&#x00026;journal=2009+IEEE+Conference+on+Computer+Vision+and+Pattern+Recognition&#x00026;pages=248-255" target="_blank">Google Scholar</a>
Synthetic and non-synthetic anthropogenic fibers in a river under the impact of paris megacity: Sampling methodological aspects and flux estimations
<a href="https://doi.org/10.1007/s11263-009-0275-4" target="_blank">CrossRef Full Text</a> | <a href="http://scholar.google.com/scholar_lookup?author=M.+Everingham&#x00026;author=L.+Van+Gool&#x00026;author=C.+K.+Williams&#x00026;author=J.+Winn&#x00026;author=A.+Zisserman+&#x00026;publication_year=2010&#x00026;title=The+pascal+visual+object+classes+(voc)+challenge&#x00026;journal=Int.+J.+Comput.+Vis&#x00026;volume=88&#x00026;pages=303-338" target="_blank">Google Scholar</a>
Automatic detection and quantification of floating marine macro-litter in aerial images: Introducing a novel deep learning approach connected to a web application in R
<a href="https://doi.org/10.1016/j.watres.2022.118902" target="_blank">CrossRef Full Text</a> | <a href="http://scholar.google.com/scholar_lookup?author=N.+Gnann&#x00026;author=B.+Baschek&#x00026;author=T.+Ternes+&#x00026;publication_year=2022&#x00026;title=Close-range+remote+sensing-based+detection+and+identification+of+macroplastics+on+water+assisted+by+artificial+intelligence%3A+a+review&#x00026;journal=Water+Res&#x00026;volume=222&#x00026;pages=118902" target="_blank">Google Scholar</a>
<a href="http://scholar.google.com/scholar_lookup?author=Y.+Guo&#x00026;author=Y.+Li&#x00026;author=L.+Wang&#x00026;author=T.+Rosing+&#x00026;publication_year=2020&#x00026;title=&#x0201C;Adafilter%3A+adaptive+filter+fine-tuning+for+deep+transfer+learning,&#x0201D;&#x00026;journal=Proceedings+of+the+AAAI+Conference+on+Artificial+Intelligence,+Vol.+34&#x00026;pages=4060-4066" target="_blank">Google Scholar</a>
<a href="http://scholar.google.com/scholar_lookup?author=Z.+Hadi&#x00026;author=N.+Sulaiman&#x00026;author=I.+A.+Halin&#x00026;author=N.+A.+M.+Yunus+&#x00026;publication_year=2016&#x00026;title=&#x0201C;Implementation+of+image+enhancement+techniques+based+on+intel+edison+platform,&#x0201D;&#x00026;journal=2016+1st+International+Conference+on+Information+Technology,+Information+Systems+and+Electrical+Engineering+(ICITISEE)&#x00026;pages=17-20" target="_blank">Google Scholar</a>
<a href="http://scholar.google.com/scholar_lookup?author=K.+He&#x00026;author=X.+Zhang&#x00026;author=S.+Ren&#x00026;author=J.+Sun+&#x00026;publication_year=2016&#x00026;title=&#x0201C;Deep+residual+learning+for+image+recognition,&#x0201D;&#x00026;journal=Proceedings+of+the+IEEE+Conference+on+Computer+Vision+and+Pattern+Recognition&#x00026;pages=770-778" target="_blank">Google Scholar</a>
<a href="http://scholar.google.com/scholar_lookup?author=G.+Huang&#x00026;author=Z.+Liu&#x00026;author=L.+Van+Der+Maaten&#x00026;author=K.+Q.+Weinberger+&#x00026;publication_year=2017&#x00026;title=&#x0201C;Densely+connected+convolutional+networks,&#x0201D;&#x00026;journal=Proceedings+of+the+IEEE+Conference+on+Computer+Vision+and+Pattern+Recognition&#x00026;pages=4700-4708" target="_blank">Google Scholar</a>
<a href="https://doi.org/10.48550/arXiv.1602.07360" target="_blank">CrossRef Full Text</a> | <a href="http://scholar.google.com/scholar_lookup?author=F.+N.+Iandola&#x00026;author=S.+Han&#x00026;author=M.+W.+Moskewicz&#x00026;author=K.+Ashraf&#x00026;author=W.+J.+Dally&#x00026;author=K.+Keutzer+&#x00026;publication_year=2016&#x00026;title=Squeezenet%3A+Alexnet-level+accuracy+with+50x+fewer+parameters+and+&#x0003C;0.5+mb+model+size&#x00026;journal=arXiv" target="_blank">Google Scholar</a>
<a href="https://doi.org/10.3390/rs12091515" target="_blank">CrossRef Full Text</a> | <a href="http://scholar.google.com/scholar_lookup?author=G.+Jakovljevic&#x00026;author=M.+Govedarica&#x00026;author=F.+Alvarez-Taboada+&#x00026;publication_year=2020&#x00026;title=A+deep+learning+model+for+automatic+plastic+mapping+using+unmanned+aerial+vehicle+(uav)+data&#x00026;journal=Remote+Sens&#x00026;volume=12&#x00026;pages=1515" target="_blank">Google Scholar</a>
Evidence that the great pacific garbage patch is rapidly accumulating plastic
The danube so colourful: a potpourri of plastic litter outnumbers fish larvae in europe&#x00027;s second largest river
<a href="https://doi.org/10.1038/nature14539" target="_blank">CrossRef Full Text</a> | <a href="http://scholar.google.com/scholar_lookup?author=Y.+LeCun&#x00026;author=Y.+Bengio&#x00026;author=G.+Hinton+&#x00026;publication_year=2015&#x00026;title=Deep+learning&#x00026;journal=Nature&#x00026;volume=521&#x00026;pages=436-444" target="_blank">Google Scholar</a>
Improved yolo based detection algorithm for floating debris in waterway
<a href="http://scholar.google.com/scholar_lookup?author=T.+-Y.+Lin&#x00026;author=M.+Maire&#x00026;author=S.+Belongie&#x00026;author=J.+Hays&#x00026;author=P.+Perona&#x00026;author=D.+Ramanan+&#x00026;publication_year=2014&#x00026;title=&#x0201C;Microsoft+coco%3A+common+objects+in+context,&#x0201D;&#x00026;journal=European+Conference+on+Computer+Vision&#x00026;pages=740-755" target="_blank">Google Scholar</a>
<a href="https://doi.org/10.3390/app11125644" target="_blank">CrossRef Full Text</a> | <a href="http://scholar.google.com/scholar_lookup?author=I.+Marin&#x00026;author=S.+Mladenovi&#x00107;&#x00026;author=S.+Gotovac&#x00026;author=G.+Zaharija+&#x00026;publication_year=2021&#x00026;title=Deep-feature-based+approach+to+marine+debris+classification&#x00026;journal=Appl.+Sci&#x00026;volume=11&#x00026;pages=5644" target="_blank">Google Scholar</a>
More than 1000 rivers account for 80% of global riverine plastic emissions into the ocean
<a href="https://doi.org/10.1109/TKDE.2009.191" target="_blank">CrossRef Full Text</a> | <a href="http://scholar.google.com/scholar_lookup?author=S.+J.+Pan&#x00026;author=Q.+Yang+&#x00026;publication_year=2009&#x00026;title=A+survey+on+transfer+learning&#x00026;journal=IEEE+Trans.+Knowl.+Data+Eng&#x00026;volume=22&#x00026;pages=1345-1359" target="_blank">Google Scholar</a>
<a href="https://doi.org/10.11591/eei.v10i5.3062" target="_blank">CrossRef Full Text</a> | <a href="http://scholar.google.com/scholar_lookup?author=F.+F.+Putra&#x00026;author=Y.+D.+Prabowo+&#x00026;publication_year=2021&#x00026;title=Low+resource+deep+learning+to+detect+waste+intensity+in+the+river+flow&#x00026;journal=Bull.+Elect.+Eng.+Inf&#x00026;volume=10&#x00026;pages=2724-2732" target="_blank">Google Scholar</a>
<a href="https://doi.org/10.48550/arXiv.1806.00451" target="_blank">CrossRef Full Text</a> | <a href="http://scholar.google.com/scholar_lookup?author=B.+Recht&#x00026;author=R.+Roelofs&#x00026;author=L.+Schmidt&#x00026;author=V.+Shankar+&#x00026;publication_year=2018" target="_blank">Google Scholar</a>
<a href="http://scholar.google.com/scholar_lookup?author=M.+Sandler&#x00026;author=A.+Howard&#x00026;author=M.+Zhu&#x00026;author=A.+Zhmoginov&#x00026;author=L.+-C.+Chen+&#x00026;publication_year=2018&#x00026;title=&#x0201C;Mobilenetv2%3A+inverted+residuals+and+linear+bottlenecks,&#x0201D;&#x00026;journal=Proceedings+of+the+IEEE+Conference+on+Computer+Vision+and+Pattern+Recognition&#x00026;pages=4510-4520" target="_blank">Google Scholar</a>
<a href="https://doi.org/10.1186/s40537-019-0197-0" target="_blank">CrossRef Full Text</a> | <a href="http://scholar.google.com/scholar_lookup?author=C.+Shorten&#x00026;author=T.+M.+Khoshgoftaar+&#x00026;publication_year=2019&#x00026;title=A+survey+on+image+data+augmentation+for+deep+learning&#x00026;journal=J.+Big+Data&#x00026;volume=6&#x00026;pages=1-48" target="_blank">Google Scholar</a>
<a href="http://scholar.google.com/scholar_lookup?author=B.+Singh&#x00026;author=L.+S.+Davis+&#x00026;publication_year=2018&#x00026;title=&#x0201C;An+analysis+of+scale+invariance+in+object+detection+snip,&#x0201D;&#x00026;journal=Proceedings+of+the+IEEE+Conference+on+Computer+Vision+and+Pattern+Recognition&#x00026;pages=3578-3587" target="_blank">Google Scholar</a>
<a href="http://scholar.google.com/scholar_lookup?author=C.+Szegedy&#x00026;author=V.+Vanhoucke&#x00026;author=S.+Ioffe&#x00026;author=J.+Shlens&#x00026;author=Z.+Wojna+&#x00026;publication_year=2016&#x00026;title=&#x0201C;Rethinking+the+inception+architecture+for+computer+vision,&#x0201D;&#x00026;journal=Proceedings+of+the+IEEE+Conference+on+Computer+Vision+and+Pattern+Recognition&#x00026;pages=2818-2826" target="_blank">Google Scholar</a>
Amsterdam urban water system as entry point of river plastic pollution
<a href="http://scholar.google.com/scholar_lookup?author=M.+Tharani&#x00026;author=A.+W.+Amin&#x00026;author=F.+Rasool&#x00026;author=M.+Maaz&#x00026;author=M.+Taj&#x00026;author=A.+Muhammad+&#x00026;publication_year=2021&#x00026;title=&#x0201C;Trash+detection+on+water+channels,&#x0201D;&#x00026;journal=International+Conference+on+Neural+Information+Processing&#x00026;pages=379-389" target="_blank">Google Scholar</a>
<a href="http://scholar.google.com/scholar_lookup?author=J.+P.+Q.+Tomas&#x00026;author=M.+N.+D.+Celis&#x00026;author=B.+T.+K.+Chan&#x00026;author=J.+A.+Flore+&#x00026;publication_year=2022&#x00026;title=&#x0201C;Trash+detection+for+computer+vision+using+scaled-yolov4+on+water+surface,&#x0201D;&#x00026;journal=The+11th+International+Conference+on+Informatics,+Environment,+Energy+and+Applications&#x00026;pages=1-8" target="_blank">Google Scholar</a>
<a href="https://doi.org/10.1088/1748-9326/ab5468" target="_blank">CrossRef Full Text</a> | <a href="http://scholar.google.com/scholar_lookup?author=C.+V.+van+Calcar&#x00026;author=T.+V.+van+Emmerik+&#x00026;publication_year=2019&#x00026;title=Abundance+of+plastic+debris+across+european+and+asian+rivers&#x00026;journal=Environ.+Res.+Lett&#x00026;volume=14&#x00026;pages=124051" target="_blank">Google Scholar</a>
<a href="https://doi.org/10.3389/frwa.2021.786936" target="_blank">CrossRef Full Text</a> | <a href="http://scholar.google.com/scholar_lookup?author=T.+van+Emmerik&#x00026;author=Y.+Mellink&#x00026;author=R.+Hauk&#x00026;author=K.+Waldschl&#x000E4;ger&#x00026;author=L.+Schreyers+&#x00026;publication_year=2022&#x00026;title=Rivers+as+plastic+reservoirs&#x00026;journal=Front.+Water&#x00026;volume=3&#x00026;pages=212" target="_blank">Google Scholar</a>
<a href="https://doi.org/10.1029/2019EA000960" target="_blank">CrossRef Full Text</a> | <a href="http://scholar.google.com/scholar_lookup?author=C.+van+Lieshout&#x00026;author=K.+van+Oeveren&#x00026;author=T.+van+Emmerik&#x00026;author=E.+Postma+&#x00026;publication_year=2020&#x00026;title=Automated+river+plastic+monitoring+using+deep+learning+and+cameras&#x00026;journal=Earth+Space+Sci" target="_blank">Google Scholar</a>
Limited long-distance transport of plastic pollution by the orange-vaal river system
<a href="http://scholar.google.com/scholar_lookup?author=N.+Wojke&#x00026;author=A.+Bewley&#x00026;author=D.+Paulus+&#x00026;publication_year=2017&#x00026;title=&#x0201C;Simple+online+and+realtime+tracking+with+a+deep+association+metric,&#x0201D;&#x00026;journal=2017+IEEE+International+Conference+on+Image+Processing+(ICIP)&#x00026;pages=3645-3649" target="_blank">Google Scholar</a>
<a href="https://doi.org/10.48550/arXiv.1411.1792" target="_blank">CrossRef Full Text</a> | <a href="http://scholar.google.com/scholar_lookup?author=J.+Yosinski&#x00026;author=J.+Clune&#x00026;author=Y.+Bengio&#x00026;author=H.+Lipson+&#x00026;publication_year=2014&#x00026;title=How+transferable+are+features+in+deep+neural+networks%3F&#x00026;journal=Adv.+Neural+Inf.+Process.+Syst&#x00026;volume=27&#x00026;pages=3220-3228" target="_blank">Google Scholar</a>
<a href="https://doi.org/10.1007/s11263-015-0812-2" target="_blank">CrossRef Full Text</a> | <a href="http://scholar.google.com/scholar_lookup?author=X.+Zhu&#x00026;author=C.+Vondrick&#x00026;author=C.+C.+Fowlkes&#x00026;author=D.+Ramanan+&#x00026;publication_year=2016&#x00026;title=Do+we+need+more+training+data%3F&#x00026;journal=Int.+J.+Comput.+Vis&#x00026;volume=119&#x00026;pages=76-92" target="_blank">Google Scholar</a>
Kapelan Z and Taormina R (2023) Advancing deep learning-based detection of floating litter using a novel open dataset
Received: 21 September 2023; Accepted: 14 November 2023; Published: 06 December 2023
Copyright &#x000A9; 2023 Jia, Vallendar, de Vries, Kapelan and Taormina. This is an open-access article distributed under the terms of the <a rel="license" href="http://creativecommons.org/licenses/by/4.0/" target="_blank">Creative Commons Attribution License (CC BY)</a>
distribution or reproduction in other forums is permitted
provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited
in accordance with accepted academic practice
distribution or reproduction is permitted which does not comply with these terms
&#x0002A;Correspondence: Tianlong Jia, <a id="encmail">dC5qaWFAdHVkZWxmdC5ubA==</a>
&#x02020;These authors have contributed equally to this work and share first authorship
Disclaimer:  All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations
Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher
94% of researchers rate our articles as excellent or goodLearn more about the work of our research integrity team to safeguard the quality of each article we publish
<a tabindex="4" href="/directory/?d=departments">Browse Departments &amp; Services</a>
<a href="/kogod/about/index.cfm" class="red-arrow red-active">About Kogod</a>
<a href="https://kogod.biz/2nzwnaf" class="red-arrow">Experience Kogod</a>
<a data-toggle="collapse" data-target="#nav4488913" href="#nav4488913" aria-expanded="false" aria-controls="nav4488913" id="ctl4488913" class="collapsed">Degrees & Programs</a>
<a data-toggle="collapse" data-target="#nav4488986" href="#nav4488986" aria-expanded="false" aria-controls="nav4488986" id="ctl4488986" class="collapsed">Admissions</a>
<a href="/kogod/students/career/index.cfm" class="red-arrow">Career Development</a>
<a href="/kogod/research/index.cfm" class="red-arrow">Research</a>
<a href="/kogod/students/index.cfm" class="red-arrow">Current Students</a>
<a data-toggle="collapse" data-target="#nav4559099" href="#nav4559099" aria-expanded="false" aria-controls="nav4559099" id="ctl4559099" class="collapsed">Faculty & Staff</a>
<a href="/kogod/news/index.cfm" class="red-arrow">News</a>
<a data-toggle="collapse" data-target="#nav5091944" href="#nav5091944" aria-expanded="false" aria-controls="nav5091944" id="ctl5091944" class="collapsed">Alumni</a>
<a href="mailto:kogod@american.edu">kogod@american.edu</a>
<a href="https://www.google.com/maps/place/KSB+-+Kogod+School+of+Business">KSB - Kogod School of Business  on a map</a>
<a href="#main-container" class="btn btn-default">Back to top </a>
Jason Javier, <a href="https://www.american.edu/kogod/undergraduate/bs-accounting.cfm" >BSA ‘19</a>, takes experiential learning to the next level. He recently accepted a position with <a href="https://www.ey.com/en_gl" >EY</a> in Berlin
strengthening his global perspective full-time within the “Big Four” accounting firm
It’s essential to understand his passion for new cultures and languages to also understand his journey at American University
After taking a German course in his first year
Javier knew he had to experience the country’s culture up-close
In spring 2018, he spent the semester studying at the <a href="https://www.whu.edu/en/" >WHU – Otto Beisheim School of Management</a>
Javier said was initially hesitant to apply because of its competitive reputation
“if you have the capability and if you study the language
a town of about 8,000 people ninety minutes between both Frankfurt and Cologne
he said the small town feel “was something I’ve never experienced.”
He relished Vallendar’s multinational culture
being exposed to a new community of industry leaders
and learning from business students from across the globe
as well as life in the professional world in DC
Javier began an internship with EY in Tyson’s Corner
Virginia—but it wasn’t long before he was called back to Germany
He joined the company’s Global Student Experience program
which placed him at one of EY’s branches in Berlin
“I never thought accounting and German would go together,” said Javier
his experience at EY and with German work culture was defining
and experienced a completely new approach to business
He was particularly struck by how transparent his supervisors were with the difference he was making in the company
“I felt like I was adding value to the team
After another highly selective interview process
Javier accepted a full-time position with EY in Berlin
now Javier hopes to bring American perspectives to Germany
and to continue to identify ways for each country to help the triple bottom line of people
He ultimately hopes to affect change in the workplace by becoming a Certified Public Accountant
helping others plan and reach their financial goals
Learn more about Kogod’s <a href="https://www.american.edu/kogod/undergraduate/bs-accounting.cfm" >bachelor of science in accounting</a> program
A business degree can be a powerful tool in today’s workforce
From Apple CEO Tim Cook to Alphabet CEO Sundar Pichai
some of the most notable names in today’s workforce gained an edge by pursuing an education in business
Armed with the critical tools and strategies to lift their ideas off the ground
they’re spearheading innovations that have altered modern society as we know it
Business school does more than just elevate aspiring entrepreneurs
however — it can lead to advances that change our world for the better
all business schools strive to continuously adapt their curriculums to meet business needs and challenges
From climate change to social and ethical concerns
business schools are offering an education that prepares graduates to address such issues
The best business degrees take this one step further: by grounding students in a globalised mindset
Graduates from these programmes have the chance to partake in a variety of experiences that grants them real-world exposure on an international scale
business-ready graduates with the insights and experiences to inspire real
Here are four business schools to save the world:
Source: WHU – Otto Beisheim School of Management
Take the <a href="https://www.whu.edu/en/programs/master-of-science-programs/master-in-finance/" target="_blank" rel="noopener">Master in Finance</a>
this programme immerses students in the finance industry through a comprehensive but flexible curriculum
What sets the WHU programme apart is the opportunity to gain international insight into the field
Outside of their usual classes in Vallendar
students can participate in a semester abroad
Those looking to gain practical experience can join internships
taking advantage of WHU’s broad network of professionals and industry contacts to do so
Ambitious learners can take this one step further by pursuing a <a href="https://www.whu.edu/en/programs/master-of-science-programs/double-degree/" target="_blank" rel="noopener">double degree</a>
students can take two semesters at WHU and two more at a partner university
after which they will earn two degree qualifications
The Part-time <a href="https://www.whu.edu/en/programs/master-of-science-programs/master-in-management/" target="_blank" rel="noopener">Master in Management</a>
is another top-ranked programme in Germany that turns learners into experts in general management
It gives working professionals the rare chance to pursue a Master of Arts whilst being employed
It doesn’t sacrifice all that makes a WHU education stand out
guaranteeing exposure to international contexts
a focus on practical tasks and forming connections with future entrepreneurs and business leaders
Throw in small classes that ensure personalised guidance and instruction and it’s no wonder why hundreds of students across the world are choosing WHU. Its focus on <a href="https://www.whu.edu/en/about-whu/diversity/" target="_blank" rel="noopener">diversity, inclusivity and “cosmopoliteness”</a> is just as influential in bringing together bright minds from every corner of the globe
While Cergy, France, is known for its popular attractions, that’s not all the city is unique for. Though a little town, the city is home to one of the top business schools in the world. The <a href="https://www.essec.edu/en/" target="_blank" rel="noopener">ESSEC Business School</a> focuses on educating students at all levels
from undergraduate to postgraduate and doctoral
The school aims to nurture future business leaders who are ready to take on careers that will benefit society
Its <a href="https://www.essec.edu/en/program/mbas/global-mba/" target="_blank" rel="noopener">Global MBA programme</a> <a href="https://www.essec.edu/en/news/essecs-global-mba-ranked-among-top-100-mbas-financial-times-ranking-and-advanced-3-places/#:~:text=News-,ESSEC&#039;s%20Global%20MBA%20ranked%20among%20the%20top%20100%20MBAs%20in,Europe%2C%20and%204th%20in%20France." target="_blank" rel="noopener">ranks #93 globally</a>
This programme is designed to focus on business concepts and international markets
It isn’t just about theory — students will gain first-hand experience through immersion in real-world projects and company visits
While the main part of the programme will be conducted on the Cergy campus
students will get the opportunity to either travel to Singapore for five weeks or one week in an emerging market or multiple other countries
this intensive programme will give students the knowledge and experience to boost their careers
This is thanks to its varied and challenging content that covers Luxury Brand Management
giving students a macro view of core business functions coupled with industry-specific skills and knowledge to help them successfully switch careers to industries such as consulting
Find out more about its exceptional MBA programme <a href="https://www.essec.edu/en/program/mbas/global-mba/" target="_blank" rel="noopener">here</a>
a Warwick postgraduate education ensures you are prepared for every aspect of it
you will get involved in MSc and MBA group and individual projects
you will also have the opportunity to participate in one of the LeadershipPlus Projects with an external corporate
Every postgraduate student draws from a wealth of human capital resources
from administrative staff and faculty members to peers and industry partners
You stand to learn from established leaders in the Change Maker Speaker Series
which has hosted the likes of Barclays Ireland CEO Kevin Wall and NASA Chief Technologist Dr
Here is a school that takes pride in producing change-makers through personal transformation and thought leadership in an inspirational and diverse applied learning environment. This spirit extends to its virtual offerings; the Warwick Business School’s MBA is <a href="https://rankings.ft.com/schools/213/warwick-business-school/rankings/2838/online-mba-ranking-2020/ranking-data" target="_blank" rel="noopener">ranked the top distance learning MBA</a> in the Financial Times Online MBA Ranking 2020
Source: Darden School of Business – University of Virginia Facebook
Preparing responsible leaders to lead the global business landscape through social transformation — that’s what the <a href="https://www.darden.virginia.edu/" target="_blank" rel="noopener">Darden School of Business</a> aims to do
it opened as the graduate business school of the Thomas Jefferson-founded University of Virginia and has since become one of the top institutions in the nation to offer a career-focused postgraduate education for ambitious individuals
Darden’s <a href="https://www.darden.virginia.edu/mba" target="_blank" rel="noopener">Full-Time MBA</a> is a 21-month taught programme that gives students unmatched access to more than 80 of the most innovative and influential business thought leaders
It begins with a core curriculum that offers a rigorous
case-based learning experience in the first year
The second year makes room for personalised interests alongside real-world exposure
where students work on consulting projects
Faculty at Darden are some of the best thinkers in the world
educating students whilst pursuing research that advances the nature of business
They are supported by a range of research centres and initiatives to progress knowledge about the field
enrich student experiences and convene forums with academics and executives in the name of promoting business development
What follows is a cohort of business leaders and executives who have the grit and tenacity to succeed in a competitive environment
They take their first steps at Darden — and graduate with the expertise to change the world
*Some of the institutions featured in this article are commercial partners of Study International
This website is using a security service to protect itself from online attacks
The action you just performed triggered the security solution
There are several actions that could trigger this block including submitting a certain word or phrase
You can email the site owner to let them know you were blocked
Please include what you were doing when this page came up and the Cloudflare Ray ID found at the bottom of this page
Christmas Song Aims to Inspire Peace in the Holy Land
Pope&#8217;s Message of Hope Launched into Space to Orbit Earth
Here is Pope Francis’ Schedule for World Youth Day 2023 in Lisbon
Marian Apparitions Must Always Point to Jesus
Germany (CNS) &#8212; The German Catholic bishops&#8217; conference and the council of the Evangelical Church in Germany plan to continue their discussions about shared Communion
German Catholic and Protestant theologians and bishops had published an appraisal of the topic in May
and it was scheduled to be discussed at the German bishops&#8217; fall plenary assembly in Fulda at the end of September
the Vatican Congregation for the Doctrine of the Faith voiced strong objections to the appraisal
saying that differences between Catholics and Protestants in the understanding of the Eucharist and the ministry were &#8220;still so grave&#8221; that they ruled out the attendance of at each other&#8217;s services
leaders of both churches identified questions that &#8220;still need to be clarified&#8221; and addressed by Catholic and Protestant sides in different ways
the open questions are so weighty that it does not feel able to allow mutual participation in general before they are clarified
especially since the question of the unity of the Catholic Church is affected here as well,&#8221; said the statement from the church leaders
Germany has many mixed marriages &#8212; Catholic and Protestant &#8212; and the issue of being able to receive Communion at each other&#8217;s churches has long been an issue of concern
one Catholic-Lutheran couple who had been married 24 years told Catholic News Service that many couples like themselves &#8220;have experienced rebuffs and coldheartedness from the churches
which has strained their marriage and led to estrangement from church.&#8221;
&#8220;It would be totally unimaginable to us to be separated from each other at the table of the Lord
when we share our whole life together in everything else,&#8221; said Herbert and Ines Heinecke
When Pope Francis visited a Lutheran parish in Rome in November 2015
a Lutheran woman married to a Catholic asked Pope Francis why she could not receive Communion when she went to Mass with her husband
The pope responded that he could not issue a general rule on shared Communion
study and then act according to their consciences
Pope Francis asked the bishops of Germany to continue working together to find broader consensus on guidelines for allowing a Protestant married to a Catholic to receive the Eucharist
The Vatican said the German bishops were told
&#8220;Pope Francis appreciates the ecumenical commitment of the German bishops and asks them to find
a result as unanimously as possible.&#8221;
CatholicPhilly.com works to strengthen the connections between people
families and communities every day by delivering the news people need to know about the Catholic Church
you and hundreds of other people become part of our mission to inform
form in the Catholic faith and inspire the thousands of readers who visit every month
Please join in the church's vital mission of communications by offering a gift in whatever amount that you can ― a single gift of $40
Your gift will strengthen the fabric of our entire Catholic community and sustain CatholicPhilly.com as your trusted news source
PREVIOUS: Pope calls for politics to renew focus on the common good
NEXT: Eventual coronavirus vaccine should belong to everyone
USCCB: Statement of USCCB on Vatican’s Document Addressing Pastoral Blessings
Pope Francis’ May 2023 Prayer Intention | Watch Video
Catholic Charities of Philadelphia Returns to Roots
Dan Tarrant Brings God to the World Through Catholic Filmmaking
<a rel="" target="_self" class="styles-module--link--6cacf breadcrumb" href="/en/">Home</a>Comillas teams shine in EuropeComillas teams shine in EuropeComillas competed in various sports disciplines at the World InterUniversities Championships (Paris) and the WHU Euromasters (Vallendar
The Comillas women&#x27;s basketball team was one of the participants
Month of international competitions for Comillas Pontifical University, with the participation of its official teams in the <a href="https://www.wiusc.com/">8th World InterUniversities Championships</a>, held in Paris, and the <a href="https://www.whu-euromasters.com/">WHU Euromasters 2023</a>
The World InterUniversities Championships hosts official teams from 75 universities around the world
More than 3,000 university athletes participated
which included teams from 12 sporting disciplines
the seven official university teams participated: basketball (women&#x27;s)
indoor football (men&#x27;s and women&#x27;s) and rugby 7 (men&#x27;s)
three coaches and two expedition leaders from Comillas Bienestar
All the students from Comillas enjoyed the experience of participating in one of the most important inter-university sports tournaments in the world, endorsed by the <a href="https://www.fisu.net/">IUSF Inter Universities Sports Federation</a>
aims to offer our students experiences through sport
as a complement to their comprehensive education
to strengthen their sense of belonging and enrich their life on campus
Comillas also participated in the WHU Euromasters
with the official basketball team (men&#x27;s)
champions of the Rector&#x27;s Trophy for men&#x27;s indoor football and men&#x27;s football 7-a-side in the previous academic year
This is a competition created in 1997 for business schools from all over Europe
in an event hosted every year by the Otto Beisheim School of Management (Vallendar
</a>Bienestar y deporte4 March 2025<a data-link="anchor" target="_self" rel="" href="https://www.comillas.edu/en/comillas-organiza-el-xxv-torneo-deportivo-unijes-2025/" class="styles-module--title--15b9d heading-sm" tabindex="-1">Comillas organises the XXV Unijes 2025 Sports Tournament</a>The Comillas teams won the title of Global Champions in a competition in which more than 400 students from Jesuit universities participated
Comillas competed in various sports disciplines at the World InterUniversities Championships (Paris) and the WHU Euromasters (Vallendar
Month of international competitions for Comillas Pontifical University, with the participation of its official teams in the <a href=\\\"https://www.wiusc.com/\\\">8th World InterUniversities Championships</a>, held in Paris, and the <a href=\\\"https://www.whu-euromasters.com/\\\">WHU Euromasters 2023</a>
the seven official university teams participated: basketball (women's)
indoor football (men's and women's) and rugby 7 (men's)
All the students from Comillas enjoyed the experience of participating in one of the most important inter-university sports tournaments in the world, endorsed by the <a href=\\\"https://www.fisu.net/\\\">IUSF Inter Universities Sports Federation</a>
champions of the Rector's Trophy for men's indoor football and men's football 7-a-side in the previous academic year
.st1{fill-rule:evenodd;clip-rule:evenodd;fill:#2a2a2a}By <a class="byline__authorLink" href="https://www.mlive.com/staff/vlozon/posts.html" title="Von Lozon at vlozon@mlive.com">Von Lozon | vlozon@mlive.com</a>SAGINAW TWP
MI -- Officers from the Saginaw Township Police Department and residents of the township were recognized at the department&#x27;s yearly Awards Program and Memorial Service Tuesday night
The event recognizes officers that have been killed in the line of duty and officers who go &quot;beyond the normal calls and may have done something extraordinary,&quot; according to Saginaw Township Police Chief Don Pussehl
This ceremony happens during Police Memorial Week every year for the Saginaw Township Police Department
Eighteen people were recognized for their service over the past year
who has been with the department for 18 years
He was selected by his peers due to &quot;putting a lot of time and effort into working on several burglary cases in the township,&quot; Pussehl said
&quot;There were four separate groups of people committing burglaries and detective Jackson was able to get them solved and hold the suspects accountable,&quot; Pussehl said
&quot;He&#x27;s also a fire investigator and works with the fire department to investigate serious fire
The Civilian Citation Award went to Phyllis Filiczkowski
McTaggart and Titus helped the police at an armed robbery scene
They got a good description of the suspect and tried to stop her before she got away
They then relayed information to Central Dispatch
Officer Matthew Oswald and Trooper Michael Roth won the Chief&#x27;s Commendation Award for that burglary case
Officer Michael Andrews won the Lifesaving Award for saving a 65-year-old man in January
&quot;(The man) went for a walk and didn&#x27;t come back (home),&quot; Pussehl said
&quot;It was a bitter cold night in January where it was below 10 degrees
His family called the police and began searching for him
Officer Andrews found him in the Nouvel Central High School parking lot where he had fallen and was laying on the ground
and if it hasn&#x27;t been for officer Andrews locating him he could&#x27;ve died.&quot;
Use of and/or registration on any portion of this site constitutes acceptance of our <a href="https://www.mlive.com/user-agreement/">User Agreement</a>, (updated 8/1/2024) and acknowledgement of our <a href="https://www.mlive.com/privacy-policy/">Privacy Policy</a>,  and <a href="https://www.mlive.com/privacy-policy/#rights_top">Your Privacy Choices and Rights</a>  (updated 1/1/2025)
© 2025 Advance Local Media LLC. All rights reserved (<a href="https://www.mlivemediagroup.com/about/?utm_source=mlive.com&amp;utm_medium=footer-article">About Us</a>)
The material on this site may not be reproduced
except with the prior written permission of Advance Local
<a href="https://www.mlive.com/communityrules/">Community Rules</a> apply to all content you upload or otherwise submit to this site
YouTube&#x27;s privacy policy is available <a href="https://policies.google.com/privacy" rel="nofollow" title="YouTube Privacy Policy" lang="en">here</a> and YouTube&#x27;s terms of service is available <a href="https://www.youtube.com/t/terms" rel="nofollow" title="YouTube Terms of Service" lang="en">here</a>
<a href="https://www.mlive.com/privacy-policy/#oba_top">Ad Choices</a>
\n        m_gallery = \"saginaw_township_police_awards\";\n        m_gallery_id = \"17856237\";\n        m_gallery_title = \"Saginaw Township Police awards ceremony \";\n        m_gallery_blog_id = \"4473\";\n        m_gallery_creation_date = \"Tuesday
 9:25 PM\";\n        m_gallery_permalink = \"http://photos.mlive.com/4473/gallery/saginaw_township_police_awards/index.html\";\n        m_gallery_json = \"https://blog.mlive.com/photogallery/4473/17856237.json\";\n        m_gallery_pagetype = \"embed\";\n        m_gallery_type = \"photo\";\n    <\/script>\n
Gallery: Saginaw Township Police awards ceremony 
Red little spotlights moved constantly across the Húrra logo on the wall
and I could feel the relaxed vibes lingering in the air
Every seat was taken by folks enjoying pints of beer and chitchatting in the dimmed room
and everyone seemed to be looking forward to an evening of newcomer Geislar’s yet unknown tunes
The band Geislar was formed earlier this year
and has just released its debut album (it was not yet out at the time of this show)
I had no idea what to expect from this novel arrangement of respected local artists
as they had only released a single song from their album
the audience clearly expecting greatness from this band of veterans
the Iceland Music Export refers to Geislar as “the new Icelandic supergroup.” So
the stage started filling up with more and more musicians
The assembly went beyond Geislar’s six official members
as a string quartet was situated halfway on and next to the stage
I was surprised to see so many musicians on stage
curious and excited to hear how all of these artists would together
The band began by tuning their instruments
sharing a few jokes and laughs amongst each other before starting
The harmony emanating from the stage soothed the crowd
The first song was a purely instrumental piece with <a href="https://www.facebook.com/pages/%C3%93skar-Gu%C3%B0j%C3%B3nsson/455931777773101" target="_blank">Óskar Guðjónsson</a> leading on the saxophone—a great way of warming the audience up to the jazzy tunes
During the first instrumental piece I wasn’t sure if the walls of Húrra could hold such a grand combination of instruments and present Geislar’s large band at its best
Once the second song began and <a href="https://www.facebook.com/pages/Sigr%C3%AD%C3%B0ur-Thorlacius-Hei%C3%B0urspiltar/123818814169" target="_blank">Sigríður Thorlacius</a> (of <a href="http://www.hjaltalinmusic.com/" target="_blank">Hjaltalín</a> fame) launched into the first notes of “Stone Cold Stone,” it all came gloriously together
Her clear and characteristic voice filled the entire room
harmonizing pleasantly between every part of this strong ensemble
keyboard and a string quartet transformed Húrra into a sparkling bubble filled with warm and soothing jazz tunes and dreamy vocals
Geislar went on to play a number of gripping tune
all of them well received by the multigenerational crowd
it felt like waking up from a dream that I didn’t want to end
<a href="https://grapevine.is/news/2025/05/05/tourist-sexually-harassed-in-blue-lagoon/">Tourist Sexually Harassed In Blue Lagoon</a>
<a href="https://grapevine.is/news/2025/05/02/tourist-held-hostage-on-may-day/">Tourist Held Hostage On May Day</a>
<a href="https://grapevine.is/news/2025/05/02/culture-night-killer-sentenced-to-prison/">Culture Night Killer Sentenced To Prison</a>
<a href="https://grapevine.is/news/2025/03/26/starbucks-to-open-in-reykjavik-soon/">Starbucks To Open In Reykjavík Soon</a>
<a href="https://grapevine.is/travel/2025/04/30/an-ode-to-flateyri/">An Ode to Flateyri</a>
<a href="https://grapevine.is/news/2025/05/02/grapevine-events-the-clock-other-grounds-film-festival-icelandic-sorcery-festival-more/">Grapevine Events: The Clock, Other-Grounds Film Festival, Icelandic Sorcery Festival &#038; More</a>
<a href="https://grapevine.is/news/2025/04/30/police-officers-spied-on-citizens-for-icelands-richest-man/">Police Officers Spied On Citizens For Iceland’s Richest Man</a>
<a href="https://grapevine.is/mag/2025/04/26/from-army-brats-to-luxury-flats/">From Army Brats To Luxury Flats</a>
Five whole days of running from one venue to another
Iceland Airwaves 2014 has come to an end— and
Day 1 +2: During the first two days of the festival
The queues at some venues were ridiculously long
everyone’s &#8216;party battery’ was fully charged
Day 3: Those of you that partied a bit too hard during the first two nights
you would have first played with the thought of just staying in and having a quiet night
But then you felt guilty for not making the most of your yellow Airwaves wristband
you could miss out on a great night of music and fun
it ended up in another late night of musical wonders&#8230; and drinking
Day 4: Saturday finds you wishing you hadn&#8217;t gone out the night before
And all those shots of Brennivín and Ópal and the rest of Iceland&#8217;s liquor-oddities weren&#8217;t such a great idea- ouch
You wanted to stay in bed forever but there was NO WAY you could miss the almost compulsory Saturday night of the festival
you ended up having a massively amusing night&#8230; yet again
Day 5: Sunday approaches and you just think: &#8216;Enough
too bad Sunday is the day of the Airwaves finale show which you probably had tickets for
Most of the day was probably spent indoors watching movies or maybe some of you did have the strength to check out some off-venue gigs… judging by those I spoke to
most likely still as tired and hungover as in the morning
It is freezing cold outside and you wish you were in your bed&#8230; but hey
it&#8217;s the Flaming Lips &#8211; and no one would want to miss how they set that unusual Airwaves venue on fire
Many of us were well knackered at this point
but it is the finale of Iceland Airwaves 2014
Even if you&#8217;re normally big a party animal
at this point you would rather just sit down comfortably and watch the band perform
<a href="https://i0.wp.com/grapevine.is/wp-content/uploads/P1160306.jpg?quality=99"></a>
The hall was full of people and you expected to get crushed by the crowd or someone pouring half their beer over you while squeezing past you
it seemed like people had decided not to invade one people&#8217;s privacy in that spectacularly Icelandic fashion
Finally a show with  enough space for everyone
space to chill &#8211; even space to keep witner coats
scarves and such on the floor without worrying about anything disappearing
You would not be separated from your friends
Nobody was left standing by the &#8216;unofficial aisle&#8217;
 getting pushed around by people that are trying to get passed
You could sway and peacefully sip that beer you decided on getting after all
as you kind of have to celebrate the final night of this amazing festival
And seeing The Flaming Lips with all their lights
balloons and other colourful props was just a treat for the eyes
After an overwhelming performance by the Flaming Lips and a few beers you thought you might as well check out what else is happening in the last few hours of Airwaves
You head to Húrra with your friends to hear the Zebra Katz play the last few tunes of Iceland Airwaves 2014
And at some point you realize: It&#8217;s time to go to sleep
Your walk back home is surprisingly quiet compared to the last four nights
Some of you might have been clever enough to take Monday off
Those Airwaves venues: they&#8217;re scattered around the downtown Reykjavík area
just a few minutes walking distance from one another
hailing or you have no idea what you want to see
The Harpa concert hall has been part of  Iceland Airwaves  for three years now
The entire building dedicates three rooms to Airwaves: Kaldalón
songwriter and guitarist Anna Calvi at Silfurberg
But if you’re more in the mood to listen to Vök’s low key electronic melodies and singing
all you have to do is walk to the next room
Harpa gives you options all night long and you don’t have to leave the building and deal with the weather outside
I met an older Airwaves veteran couple (attending for their tenth year now!) at Húrra on Thursday during the Ojba Rasta set
They did not seem to be feeling the reggae beats
They enjoy music and have been for the past ten years
they have no interest in standing around in long queues outside
or arriving at gigs and realize the music sucks
Last year we spent the entire Airwaves at Harpa and it was great
There will always be something we can or want to see in one of the three rooms.” They had spent the whole of Wednesday night at Harpa and decided to be more adventurous on Thursday by trying a few different venues
We feel a bit out of place here and we are too old to jump from venue to venue
It is too chaotic for us,&#8221; they said
noting they would spend the entire next day at Harpa
So it seems like Harpa is not only the place to be when the weather is bad or you have no clue what you want to see.It is also for those Airwaves veterans that have just come to enjoy a more sophisticated night of music
.st1{fill-rule:evenodd;clip-rule:evenodd;fill:#2a2a2a}By <a class="byline__authorLink" href="https://www.silive.com/staff/mgrunlun/posts.html" title="Maura Grunlund at mauragrunlund@siadvance.com">Maura Grunlund | mauragrunlund@siadvance.com</a>Staten Island Advance/Irving SilversteinOutside the shrine
Sister Mary Angela Macari welcomes the visitors disembarking from their bus.STATEN ISLAND
- More than 50 pilgrims from Puerto Rico visited the Schoenstatt Shrine of Light in West Brighton in anticipation of a jubilee celebration
While the shrine isn’t well known among Staten Islanders
it is part of the worldwide Schoenstatt Apostolic Movement that has a strong following in Puerto Rico
it is one of only nine such shrines in the U.S
About 200 identical shrines are located worldwide
The Schoenstatt movement was founded in 1914 shortly after the start of WWI by the Rev
Father Kentenich was interned in Dachau concentration camp from 1941 to 1945 and is on the path to sainthood in the Roman Catholic Church
some New Yorkers who originally were from Puerto Rico heard about the Schoenstatt Apostolic Movement while attending their parishes in Brooklyn
The priest returned the favor when he visited the movement members on Sept
established in that same year on Castleton Avenue
Some of those who met Father Kentenich in Milwaukee later returned to their native land and established the movement in Puerto Rico
52 pilgrims — including four who made the trip to Milwaukee in 1962—and two Schoenstatt sisters flew to New York as a culmination and high point of their golden jubilee pilgrimage
The group had visited Milwaukee prior to seeing the Island
The pilgrims were “following in the footsteps” of not only Father Kentenich but the original ones who brought the movement to Puerto Rico
The pilgrims left the small island of Puerto Rico
in many cases to live in Brooklyn and work in Manhattan
“Schoenstatt spirituality gave them encouragement and assistance in their everyday life,” Sister Macari said of them
While retracing their founding heritage here in the U.S.
the pilgrims visited the Schoenstatt Sisters of Mary house in West Brighton
The pilgrims then headed immediately into the shrine
the pilgrims also walked to the previous Schoenstatt house on Castleton Avenue
listened to a recording of Father Kentenich speaking
Church in Brooklyn and went sightseeing in Manhattan
The West Brighton shrine has been located at its current location
also living in the West Brighton house are Sisters M
the West Brighton shrine is unique as it is inside the house and not in a separate chapel
with a Madonna and Child painting of the Mother Thrice Admirable
Saint Paul and Saint Joseph remind people of the intercession of holy people in heaven who join their prayers with those of the visitors,” Sister Macari said
reminder of the loving fatherly care of God
and a silver and gold dove suspended from the ceiling symbolizes the constant presence and power of the Holy Spirit
The Blessed Sacrament is reserved in the tabernacle as the focus of the prayers demonstrating that Mary leads everyone to her Son.”
Use of and/or registration on any portion of this site constitutes acceptance of our <a href="https://www.silive.com/user-agreement/">User Agreement</a>, (updated 8/1/2024) and acknowledgement of our <a href="https://www.silive.com/privacy-policy/">Privacy Policy</a>,  and <a href="https://www.silive.com/privacy-policy/#rights_top">Your Privacy Choices and Rights</a>  (updated 1/1/2025)
© 2025 Advance Local Media LLC. All rights reserved (<a href="https://www.statenislandmediagroup.com/about-us/">About Us</a>)
<a href="https://www.silive.com/communityrules/">Community Rules</a> apply to all content you upload or otherwise submit to this site
<a href="https://www.silive.com/privacy-policy/#oba_top">Ad Choices</a>
.st1{fill-rule:evenodd;clip-rule:evenodd;fill:#2a2a2a}By <a class="byline__authorLink" href="https://www.silive.com/staff/sisiadvance/posts.html" title="Staten Island Advance">Staten Island Advance</a>Advance File PhotoThe  Schoenstatt Shrine of Light and activites by the Schoenstatt Sisters of Mary
both based at 337 Cary Avenue.STATEN ISLAND
-- A hidden gem in West Brighton is the Schoenstatt Shrine of Light
which averages about 200 faithful visitors per month
About 200 shrines are located throughout the world for the Roman Catholic Schoenstatt Apostolic Movement
As one of nine Schoenstatt Shrines in the United States
the annual May Crowning at the borough shrine typically attracts more than 450 people
from as far north as Massachusetts to as far south as Washington
The original Schoenstatt shrine is located in a chapel dating from the 1400s
in the town of Vallendar in the Schoenstatt valley near Koblenz
The movement was founded in that chapel in 1914
He was interned in Dachau concentration camp from 1941-1945
and is on the path to sainthood in the Roman Catholic Church
Each of the 200 Schoenstatt shrines look identical inside -- they feature a baroque wood-carved altar; a painting of Madonna and Child; statues of saints Michael
Paul and Joseph; an eye symbol representing God
Father Kentenich visited Staten Island and its Schoenstatt Sisters of Mary at their first convent on Castleton Avenue
Today -- and since 1977 -- the shrine is located at 337 Cary Ave
on the anniversary of Father Kentenich&#x27;s visit
the faithful gathered at the shrine to pray for the victims of 9/11 and their families
The Schoenstatt Shrine Light is open daily from 8 a.m. to 8 p.m. For more information, call 718-727-8005, or email Sister Angela Macari at <a href="https://blog.advance.net/mt-static/html/shrineNY@schsrsmary.org">shrineNY@schsrsmary.org</a>