MARVEL: Multimodal Extreme Scale Data Analytics for Smart Cities Environments

Bajović, Dragana; Bakhtiarnia, Arian; Bravos, George; Brutti, Alessio; Burkhardt, Felix; Cauchi, Daniel; Chazapis, Antony; Cianco, Claire; Dall’Asen, Nicola; Delić, Vlado; Dimou, Christos; Djokic, Djordje; Escobar-Molero, Antonio; Esterle, Lukas; Eyben, Florian; Farella, Elisabetta; Festi, Thomas; Geromitsos, Artemios; Giakoumakis, Giannis; Hatzivasilis, George; Ioannidis, Sotiris; Iosifidis, Alexandros; Kallipolitou, Theodora; Kalogiannis, Grigorios; Kiousi, Akrivi; Kopanaki, Despina; Marazakis, Manolis; Markopoulou, Stella; Muscat, Adrian; Paissan, Francesco; Lobo, Tomás Pariente; Pavlović, Dušan; Raptis, Theofanis P.; Ricci, Elisa; Saez, Borja; Sahito, Farhan; Scerri, Kenneth; Schuller, Björn W.; Simic, Nikola; Spanoudakis, George; Tomasi, Alex; Triantafyllopoulos, Andreas; Valerio, Lorenzo; Villazan, Javier; Wang, Yiming; Xuereb, André; Zammit, Johan

doi:10.1109/balkancom53780.2021.9593258

Cited by 16 publications

(9 citation statements)

References 14 publications

(9 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Visual crowd counting is the task of counting the total number of people present in a scene given an image of that scene. In the context of smart cities, visual crowd counting can be used to monitor crowds, collect crowd statistics over time, and increase safety in areas of particular interest [1]. The input to a visual crowd counting method is an image or video frame of the scene, and the output is a single number representing the total count.…”

Section: A Visual Crowd Countingmentioning

confidence: 99%

“…However, different methods will lead to different types of visual quality reduction for lowering the amount of bandwidth needed to transmit the compressed video frames. Thus, they can have a different impact on the accuracy achieved by the DNN used to solve the visual analysis task 1 . In order to achieve a good compromise between high image compression and high accuracy on the visual analysis task at hand, the effect of using different lossy image compression methods needs to be carefully investigated.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Analysis of the Effect of Low-Overhead Lossy Image Compression on the Performance of Visual Crowd Counting for Smart City Applications

Bakhtiarnia¹,

Leporowski²,

Esterle³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Images and video frames captured by cameras placed throughout smart cities are often transmitted over the network to a server to be processed by deep neural networks for various tasks. Transmission of raw images, i.e., without any form of compression, requires high bandwidth and can lead to congestion issues and delays in transmission. The use of lossy image compression techniques can reduce the quality of the images, leading to accuracy degradation. In this paper, we analyze the effect of applying low-overhead lossy image compression methods on the accuracy of visual crowd counting, and measure the trade-off between bandwidth reduction and the obtained accuracy.

show abstract

Section: A Visual Crowd Countingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Analysis of the Effect of Low-Overhead Lossy Image Compression on the Performance of Visual Crowd Counting for Smart City Applications

Bakhtiarnia¹,

Leporowski²,

Esterle³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Both parts split the images into training and test sets. Since these datasets do not provide validation data, we randomly take 20% from the training set of part A and 10% from part B as validation data 1 . SASNet [12] is the state-of-the-art DNN for crowd counting on the Shanghai Tech dataset at the time of this writing.…”

Section: B Crowd Countingmentioning

confidence: 99%

“…Many applications in smart cities, such as crowd monitoring, traffic surveillance and anomaly detection, utilize deep learning to process visual information [1], [2]. In such settings, typically the video frames taken by many cameras installed throughout the city are transmitted to a few edge or cloud servers to be processed.…”

Section: Introductionmentioning

confidence: 99%

Crowd Counting on Heavily Compressed Images with Curriculum Pre-Training

Bakhtiarnia¹,

Zhang²,

Iosifidis³

2022

Preprint

Self Cite

View full text Add to dashboard Cite

JPEG image compression algorithm is a widely used technique for image size reduction in edge and cloud computing settings. However, applying such lossy compression on images processed by deep neural networks can lead to significant accuracy degradation. Inspired by the curriculum learning paradigm, we present a novel training approach called curriculum pre-training (CPT) for crowd counting on compressed images, which alleviates the drop in accuracy resulting from lossy compression. We verify the effectiveness of our approach by extensive experiments on three crowd counting datasets, two crowd counting DNN models and various levels of compression. Our proposed training method is not overly sensitive to hyperparameters, and reduces the error, particularly for heavily compressed images, by up to 19.70%.

show abstract

“…Data stream processing is gaining more and more attention as a new implementation paradigm to manage real-time datadriven applications in numerous smart-* domains, such as cities [1], industry [2], networking [3], and agriculture [4]. This paradigm is more suitable to modern application developers that consist of vertically separated engineering teams, each one managing a loosely coupled sub-system.…”

Section: Introductionmentioning

confidence: 99%

On Efficiently Partitioning a Topic in Apache Kafka

Raptis,

Passarella

2022

Preprint

Self Cite

View full text Add to dashboard Cite

Apache Kafka addresses the general problem of delivering extreme high volume event data to diverse consumers via a publish-subscribe messaging system. It uses partitions to scale a topic across many brokers for producers to write data in parallel, and also to facilitate parallel reading of consumers. Even though Apache Kafka provides some out of the box optimizations, it does not strictly define how each topic shall be efficiently distributed into partitions. The well-formulated fine-tuning that is needed in order to improve an Apache Kafka cluster performance is still an open research problem. In this paper, we first model the Apache Kafka topic partitioning process for a given topic. Then, given the set of brokers, constraints and application requirements on throughput, OS load, replication latency and unavailability, we formulate the optimization problem of finding how many partitions are needed and show that it is computationally intractable, being an integer program. Furthermore, we propose two simple, yet efficient heuristics to solve the problem: the first tries to minimize and the second to maximize the number of brokers used in the cluster. Finally, we evaluate its performance via largescale simulations, considering as benchmarks some Apache Kafka cluster configuration recommendations provided by Microsoft and Confluent. We demonstrate that, unlike the recommendations, the proposed heuristics respect the hard constraints on replication latency and perform better w.r.t. unavailability time and OS load, using the system resources in a more prudent way.

show abstract

MARVEL: Multimodal Extreme Scale Data Analytics for Smart Cities Environments

Cited by 16 publications

References 14 publications

Analysis of the Effect of Low-Overhead Lossy Image Compression on the Performance of Visual Crowd Counting for Smart City Applications

Analysis of the Effect of Low-Overhead Lossy Image Compression on the Performance of Visual Crowd Counting for Smart City Applications

Crowd Counting on Heavily Compressed Images with Curriculum Pre-Training

On Efficiently Partitioning a Topic in Apache Kafka

Contact Info

Product

Resources

About