Design and Deployment of an Image Polarity Detector with Visual Attention

Ragusa, Edoardo; Apicella, Tommaso; Gianoglio, Christian; Zunino, Rodolfo; Gastaldo, Paolo

doi:10.1007/s12559-021-09829-6

Cited by 18 publications

(13 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This section illustrates the results of the experiments performed in the three test scenarios. According to our previous findings [38] we observed that moving the same architecture from the Desktop to the embedded platform has negligible impact on the detection accuracy, while it mostly affects speed and memory footprint. Therefore, in Sections 5.1-5.3 we first illustrate, for each scenario, our achievements using the Desktop platform and, then, in Section 5.4, we analyze the impact of deploying T-RexNet on the Jetson Nano.…”

Section: Resultssupporting

confidence: 54%

“…TensorTRT can adopt different data sizes when deploying a network: standard floating-point representation (FP32), half-precision floating point (FP16) and 8-bit integer representation (INT8). The experiments were conducted with the FP16 format since this provides a good trade-off between accuracy and power consumption [38]. In addition, the results proved that FP16 is indeed sufficient to reach good frame rates using Jetson Nano.…”

Section: Deploymentmentioning

confidence: 98%

“…Accordingly, a direct measure would yield biased results. Indeed, literature proves that similar models can be deployed in devices using a smaller memory footprint [38].…”

Section: Deployment Of T-rexnet On the Jetson Nanomentioning

confidence: 99%

See 2 more Smart Citations

T-RexNet—A Hardware-Aware Neural Network for Real-Time Detection of Small Moving Objects

Canepa

Ragusa

Zunino

et al. 2021

Sensors

Self Cite

View full text Add to dashboard Cite

This paper presents the T-RexNet approach to detect small moving objects in videos by using a deep neural network. T-RexNet combines the advantages of Single-Shot-Detectors with a specific feature-extraction network, thus overcoming the known shortcomings of Single-Shot-Detectors in detecting small objects. The deep convolutional neural network includes two parallel paths: the first path processes both the original picture, in gray-scale format, and differences between consecutive frames; in the second path, differences between a set of three consecutive frames is only handled. As compared with generic object detectors, the method limits the depth of the convolutional network to make it less sensible to high-level features and easier to train on small objects. The simple, Hardware-efficient architecture attains its highest accuracy in the presence of videos with static framing. Deploying our architecture on the NVIDIA Jetson Nano edge-device shows its suitability to embedded systems. To prove the effectiveness and general applicability of the approach, real-world tests assessed the method performances in different scenarios, namely, aerial surveillance with the WPAFB 2009 dataset, civilian surveillance using the Chinese University of Hong Kong (CUHK) Square dataset, and fast tennis-ball tracking, involving a custom dataset. Experimental results prove that T-RexNet is a valid, general solution to detect small moving objects, which outperforms in this task generic existing object-detection approaches. The method also compares favourably with application-specific approaches in terms of the accuracy vs. speed trade-off.

show abstract

Section: Resultssupporting

confidence: 54%

Section: Deploymentmentioning

confidence: 98%

See 1 more Smart Citation

T-RexNet—A Hardware-Aware Neural Network for Real-Time Detection of Small Moving Objects

Canepa

Ragusa

Zunino

et al. 2021

Sensors

Self Cite

View full text Add to dashboard Cite

show abstract

“…Beyond the comparison of these compact architectures inspired to their larger counterparts, few works that empirically study or propose different approaches to tackle the limitations posed by embedded devices have also been published in related task such as polarity detection from visual data [ 26 , 27 ], speech recognition [ 28 ] and conversational agents [ 29 ]. In addition, the trade-off between computational cost and generalization performance for sentiment analysis was studied in [ 30 , 31 ], but the focus was not on the deployment of the solution.…”

Section: Related Workmentioning

confidence: 99%

Emotion Recognition on Edge Devices: Training and Deployment

Pandelea

Ragusa

Apicella

et al. 2021

Sensors

Self Cite

View full text Add to dashboard Cite

Emotion recognition, among other natural language processing tasks, has greatly benefited from the use of large transformer models. Deploying these models on resource-constrained devices, however, is a major challenge due to their computational cost. In this paper, we show that the combination of large transformers, as high-quality feature extractors, and simple hardware-friendly classifiers based on linear separators can achieve competitive performance while allowing real-time inference and fast training. Various solutions including batch and Online Sequential Learning are analyzed. Additionally, our experiments show that latency and performance can be further improved via dimensionality reduction and pre-training, respectively. The resulting system is implemented on two types of edge device, namely an edge accelerator and two smartphones.

show abstract

“…Given the above, online cyber social networks can tell us what people are talking about right now and what is happening in the world. Especially, identifying the polarity of users in various fields on cyber social networks, has also become a popular and useful research topic (see [1,2,3,4,5,6,7,8,9,10,11]).…”

Section: Introductionmentioning

confidence: 99%

Mathematical modelling for detection of polarity in online social networks using Dempster−Shafer theory and fuzzy functions

Shekarbaigi

2022

Preprint

View full text Add to dashboard Cite

In this paper, we introduce a mathematical modeling of a relative polarity of onlinesocial networks users. For this purpose, we create a fuzzy model of communities andfinally, we detect the polarity of all users. To achieve this goal, we are faced with theproblem of uncertainty. To solve this problem, we calculated the major methods ofdecision integration, including voting, rank, score, behavior and the Dempster Shafer method. Another initiative of this paper is in it’s ultimate goal which clearly deals with the polarity of a society by a new mathematical modeling. In order to be able toevaluate our proposed method correctly and accurately, we should first implement it and then examine the tools that are produced. Python programming language has been used in Spider software to implement community polarity detection. At each stage we have used the required libraries which have about 3500 lines of useful code. To evaluate the proposed method in Section 2, we implemented our proposed method. The evaluation of the proposed method was done in two ways, first, we compared the second phase of the third phase with different methods of decision integration and after selecting the best method, we reviewed the output results according to the opinions of experts.

show abstract

Design and Deployment of an Image Polarity Detector with Visual Attention

Cited by 18 publications

References 43 publications

T-RexNet—A Hardware-Aware Neural Network for Real-Time Detection of Small Moving Objects

T-RexNet—A Hardware-Aware Neural Network for Real-Time Detection of Small Moving Objects

Emotion Recognition on Edge Devices: Training and Deployment

Mathematical modelling for detection of polarity in online social networks using Dempster−Shafer theory and fuzzy functions

Contact Info

Product

Resources

About