Septentrio Academic Publishing
Not a member yet
6415 research outputs found
Sort by
Evaluating current state of monocular 3D pose models for golf
No full textMonocular 3D human pose estimation has reached an impressive performance. State-of-the-art mod- els predict joint locations that can be accurately reprojected back into the image, resulting in vi- sually convincing detections. However, our aim is to use the predicted poses in a domain with high- frequency movements, that is, for video of ath- letes performing golf swings. Our investigation is based on accurate marker-based motion capture data. Also, for our data, the predicted 3D joint locations look convincing when we reproject them into the image. However, by quantitatively com- paring the results with the motion capture data, we see significant model errors that are too erroneous to be used for any kinematic analysis of the move- ments. Thus we conclude that the current models cannot be used out of the box for advanced golf analytics
Hybrid bayesian convolutional neural network object detection architectures for tracking small markers in automotive crashtest videos
No full textAutomotive crash tests are an important aspect of everyday safety, where accurate measurements and evaluations play a crucial role. In order to automate this process, we implement a bayesian hybrid computer vision model to detect two different kinds of target markers. These are commonly used in crash tests, by e.g. automotive manufacturers, to aid the localisation of the car’s positional data by attaching these markers at different positions to the vehicle. A tracking algorithm is subsequently used to add contextual time information to the marker objects. The extracted information can then be used in downstream tasks to calculate important metrics for the crash test evaluation, e.g. the speed, momentum, acceleration and trajectory at particular parts of the car during different stages of the crash test. The model consists of a pre-trained Faster-RCNN for the region proposals with the addition of a bayesian convolutional neural network to estimate a statistical uncertainty on the model’s classifications. This uncertainty estimation can be used as a tool to improve safety in uncertain edge cases in videos where lighting conditions and light reflections are not optimal. Our pipeline achieves an average recall and precision of 0.89 and 0.99, respectively, when applied to test data. This outperforms the recall of state of the art models like the Faster-RCNN Resnet-152 by more than 28\% while delivering slightly better precision, increasing robustness in most of the tested use-cases
Improved Imagery Throughput via Cascaded Uncertainty Pruning on U-Net++
No full textThe extensive use of machine learning inferences in real-life earth observation and remote sensing cases has grown over recent years. Network pruning has been carefully studied in various applications to speed up the machine learning workflow, but mainstream pruning strategies often focus on specific connection significancy rather than the sample difficulty. U-Net++ as a well-versed and capable semantic segmentation deep convolutional neural network architecture, as well as its equivalents, are all facing the challenge of overconfidence, which will create barriers for a robust uncertainty-based pruning strategy to be designed. In the following study, we analyzed the efficiency of deep neural networks and semantic segmentation in satellite imagery analysis, and proposed a new tailored workf low of dynamic pruning for U-Net++ by combining the ideas of network calibration and uncertainty and defining the inference complexity of network input samples. We tested and illustrated the capability of this new workflow and delivered a successful comparative study on its effectiveness on the DeepGlobe satellite imagery road extraction dataset and how it can greatly reduce the computational cost with little performance drop
Paradokser med en kommisjon?
Get PDFArtikkel i publikasjonen Om den norske sannhets- og forsoningskommisjonen: foreløpige funn
La expresión del futuro en las variedades oral y escrita del español de Uruguay
No full textIn this article, we analyze the differences in the distribution and interpretation of the synthetic future [-RÁ] and the periphrastic future [IR + a + Inf] in the oral and written records of Uruguayan Spanish. The quantitative and qualitative analysis of two corpora -oral and written- shows that the written language maintains the preference of [-RÁ] to refer to the future while in orality it is pronounced by [IR + a + Inf] and [-RÁ] is used solely as an uncertainty operator. The type of sentences in which [-RÁ] appears in orality (independent and subordinates with complete left periphery) is an indication that this operator is interpreted at the extrapropositional level of the SFuerza projection. In the geographical area studied, the two ways to refer to the future coexist, which leads us to conclude that there are two grammars, one associated with the written language and the other with the oral language.En este artículo analizamos las diferencias en la distribución y en la interpretación del futuro sintético [-RÁ] y del futuro perifrástico [IR + a + Inf] en los registros oral y escrito del español del Uruguay. El análisis cuantitativo y cualitativo de dos corpus -oral y escrito- muestra que la lengua escrita mantiene la preferencia de [-RÁ] para referir al futuro mientras en la oralidad es expresado por [IR + a + Inf] y [-RÁ] se emplea únicamente como operador de incertidumbre. El tipo de oraciones en que aparece [-RÁ] en la oralidad (independientes y subordinadas con periferia izquierda completa) constituye un indicio de que este operador se interpreta en el nivel extraproposicional de la proyección SFuerza. En la zona geográfica estudiada, conviven las dos formas para referir al futuro, lo que nos lleva a concluir que existen dos gramáticas, una asociada a la lengua escrita y otra, a la lengua oral
From product-sales models to digital ecosystems and open science: The case of scholarly journal publishers in a small language country
No full textWatch VIDEO.
Scholarly publishers in small language countries face unique challenges, such as limited funding and resources, lack of visibility and recognition, and the dominance of international commercial publishers in the global academic publishing market. Therefore, understanding the role of publishers in promoting scholarly research in small language countries is crucial. This research aims to evaluate the publication of scholarly journals in a small language country (Lithuania) by analyzing the owners and publishers of scholarly journals. After collecting information about all scholarly journals published in 2020, a list of publishers and owners was compiled. The results show that since 1990, the publishing of Lithuanian scholarly journals has grown significantly: 225 scholarly journals were published in Lithuania, by 73 different publishers, in 2020. Social publishers, mostly state-funded universities, scientific institutes, colleges, still had the largest market share of scholarly periodicals, but commercial publishers also appeared and started to make a business out of publishing scholarly journals. Judging by the number of published articles, the five largest Lithuanian social publishers – state universities – published almost half of all scholarly articles published by Lithuanian publishers. Journals of social science were published the most, and one in five journals published in Lithuania applied relatively small publication fees. The analyzed data on publishers show that Lithuanian publishers have tried to take over and reorganize a number of scholarly journals from the Soviet era, while adapting to the dynamically changing world of scholarly publishing, and moving from simple product-sales models of the 20th century to new digital ecosystems of the 21st century, in which the essential distribution of publications is carried out digitally. The emerging diversity of publishers, the relatively moderate application of publication fees, and the growth in the number of articles show that Lithuanian publishers are quite well prepared for further challenges – the implementation of the European Science Foundation Plan S recommendations
Hopping on the AI train? Ethical and practical considerations for the adoption of AI tools in research and higher education
No full textThe most recent wave of developments in AI has commanded both awe and apprehension about what these innovations mean for the future of science, scholarly communication, and librarianship. In a paradigm characterized by fierce competition, as illustrated by the pervasive culture of publish or perish, the rapid development and diffusion of AI-based tools can bring challenges to existing frameworks for research ethics. AI is likely to challenge the integrity of scientific enterprise in ways that are yet to be seen, and these impacts extend to the operation of libraries, publishers and more.
This workshop seeks to promote a collective exploration of what implications the adoption of AI tools brings to the scientific endeavor. No prior knowledge of AI is required. In advance of the workshop, the room will be arranged in three tables, each marked with one of three topics: (1) research values, accountability and integrity, (2) publishing, open science and science communication, (3) competencies of teaching and research support staff. Participants chose their seats according to their preferred topic (not necessarily related to their specific professional identity).
The workshop agenda is as follows:
Setting the stage (10 min). Welcome and agenda for the workshop. Information about the posters. The facilitators will introduce themselves and present the three themes.
research values, accountability, and integrity.
publishing, open science, and science communication.
competencies of teaching and research support staff.
Group work (50 min). According to each table’s theme, participants will be given time to work with dilemmas or case examples. Each table will work on the following perspectives:
What? Discussion: Participants discuss while writing on Post-its or directly on A2 paper sheet comments, and notes: What ethical issues are emerging from the given dilemma? Participants will be requested to arrange their notes in topic clusters (affinity mapping) and give them names.
So what? Discussion: What do these developments mean (for science, libraries, society)? What are the possible consequences? Why should anyone care? Each group will be oriented to create a mind map (https://www.sessionlab.com/methods/mind-map) of this landscape on a new A2 paper sheet.
Now what? Discussion: Do ethical frameworks need to be revised (and how)? What measures need to be in place as AI tools become mainstream? Who is in a position to implement such measures? What is the role of libraries, publishers, and other support units in safeguarding the upholding of ethical principles? The groups will be guided to use a simple impact and effort matrix pre-defined on an A2 paper sheet (https://www.sessionlab.com/methods/impact-and-effort-matrix).
Plenum discussion (30 min).
The expected outcome is that participants gain a clearer understanding of the ethical challenges to come as a consequence of AI, and more confidence to champion these discussions at their home institutions. Moreover, facilitators expect to publish insights from the workshop (in a format yet to be determined). Finally, the workshop aims to plant the seeds for future work on guidelines that libraries can use when assessing the adoption or promotion of new AI tools
