Search

Computer vision is the science related to teaching machines to see and understand digital images or videos. During the last decade, computer vision has seen tremendous progress on perception tasks such as object detection, semantic segmentation, and video action recognition, which lead to the development and improvements of important industrial applications such as self-driving cars and medical image analysis. These advances are mainly due to fast computation offered by GPUs, the development of high capacity models such as deep neural networks, and the availability of large datasets, often composed by a variety of modalities. In this thesis, we explore how multimodal data can be used to train deep convolutional neural networks. Humans perceive the world through multiple senses, and reason over the multimodal space of stimuli to act and understand the environment. One way to improve the perception capabilities of deep learning methods is to use different modalities as input, as it offers different and complementary information about the scene. Recent multimodal datasets for computer vision tasks include modalities such as depth maps, infrared, skeleton coordinates, and others, besides the traditional RGB. This thesis investigates deep learning systems that learn from multiple visual modalities. In particular, we are interested in a very practical scenario in which an input modality is missing at test time. The question we address is the following: how can we take advantage of multimodal datasets for training our model, knowing that, at test time, a modality might be missing or too noisy? The case of having access to more information at training time than at test time is referred to as learning using privileged information. In this work, we develop methods to address this challenge, with special focus on the tasks of action and object recognition, and on the modalities of depth, optical flow, and RGB, that we use for inference at test time. This thesis advances the art of multimodal learning in three different ways. First, we develop a deep learning method for video classification that is trained on RGB and depth data, and is able to hallucinate depth features and predictions at test time. Second, we build on this method and propose a more generic mechanism based on adversarial learning to learn to mimic the predictions originated by the depth modality, and is able to automatically switch from true depth features to generated depth features in case of a noisy sensor. Third, we develop a method that learns a single network trained on RGB data, that is enriched with additional supervision information from other modalities such as depth and optical flow at training time, and that outperforms an ensemble of networks trained independently on these modalities.

Submitted by Nuno Augusto Pereira Garcia (nuno@mpfoadvogados.com.br) on 2020-04-17T01:50:08Z No. of bitstreams: 1 TESE - Nuno Augusto Pereira Garcia (Erro Médico) 10-04-2020.pdf: 1215291 bytes, checksum: 6449fcf0df69a0a2779eaa7dbb218883 (MD5) Submitted by Nuno Augusto Pereira Garcia (nuno@mpfoadvogados.com.br) on 2020-04-17T01:50:08Z No. of bitstreams: 1 TESE - Nuno Augusto Pereira Garcia (Erro Médico) 10-04-2020.pdf: 1215291 bytes, checksum: 6449fcf0df69a0a2779eaa7dbb218883 (MD5) Submitted by Nuno Augusto Pereira Garcia (nuno@mpfoadvogados.com.br) on 2020-04-17T01:50:08Z No. of bitstreams: 1 TESE - Nuno Augusto Pereira Garcia (Erro Médico) 10-04-2020.pdf: 1215291 bytes, checksum: 6449fcf0df69a0a2779eaa7dbb218883 (MD5) Approved for entry into archive by ROSANGELA APARECIDA LOBO null (rosangelalobo@btu.unesp.br) on 2020-04-17T21:46:04Z (GMT) No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Submitted by Nuno Augusto Pereira Garcia (nuno@mpfoadvogados.com.br) on 2020-04-17T01:50:08Z No. of bitstreams: 1 TESE - Nuno Augusto Pereira Garcia (Erro Médico) 10-04-2020.pdf: 1215291 bytes, checksum: 6449fcf0df69a0a2779eaa7dbb218883 (MD5) Approved for entry into archive by ROSANGELA APARECIDA LOBO null (rosangelalobo@btu.unesp.br) on 2020-04-17T21:46:04Z (GMT) No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Made available in DSpace on 2020-04-17T21:46:04Z (GMT). No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Previous issue date: 2020-02-20 Made available in DSpace on 2020-04-17T21:46:04Z (GMT). No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Previous issue date: 2020-02-20 Approved for entry into archive by ROSANGELA APARECIDA LOBO null (rosangelalobo@btu.unesp.br) on 2020-04-17T21:46:04Z (GMT) No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Approved for entry into archive by ROSANGELA APARECIDA LOBO null (rosangelalobo@btu.unesp.br) on 2020-04-17T21:46:04Z (GMT) No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Made available in DSpace on 2020-04-17T21:46:04Z (GMT). No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Previous issue date: 2020-02-20 Submitted by Nuno Augusto Pereira Garcia (nuno@mpfoadvogados.com.br) on 2020-04-17T01:50:08Z No. of bitstreams: 1 TESE - Nuno Augusto Pereira Garcia (Erro Médico) 10-04-2020.pdf: 1215291 bytes, checksum: 6449fcf0df69a0a2779eaa7dbb218883 (MD5) Approved for entry into archive by ROSANGELA APARECIDA LOBO null (rosangelalobo@btu.unesp.br) on 2020-04-17T21:46:04Z (GMT) No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Made available in DSpace on 2020-04-17T21:46:04Z (GMT). No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Previous issue date: 2020-02-20 Made available in DSpace on 2020-04-17T21:46:04Z (GMT). No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Previous issue date: 2020-02-20 Made available in DSpace on 2020-04-17T21:46:04Z (GMT). No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Previous issue date: 2020-02-20 Approved for entry into archive by ROSANGELA APARECIDA LOBO null (rosangelalobo@btu.unesp.br) on 2020-04-17T21:46:04Z (GMT) No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Made available in DSpace on 2020-04-17T21:46:04Z (GMT). No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Previous issue date: 2020-02-20 Submitted by Nuno Augusto Pereira Garcia (nuno@mpfoadvogados.com.br) on 2020-04-17T01:50:08Z No. of bitstreams: 1 TESE - Nuno Augusto Pereira Garcia (Erro Médico) 10-04-2020.pdf: 1215291 bytes, checksum: 6449fcf0df69a0a2779eaa7dbb218883 (MD5) Made available in DSpace on 2020-04-17T21:46:04Z (GMT). No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Previous issue date: 2020-02-20 Approved for entry into archive by ROSANGELA APARECIDA LOBO null (rosangelalobo@btu.unesp.br) on 2020-04-17T21:46:04Z (GMT) No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Made available in DSpace on 2020-04-17T21:46:04Z (GMT). No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Previous issue date: 2020-02-20 Made available in DSpace on 2020-04-17T21:46:04Z (GMT). No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Previous issue date: 2020-02-20 Made available in DSpace on 2020-04-17T21:46:04Z (GMT). No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Previous issue date: 2020-02-20 Approved for entry into archive by ROSANGELA APARECIDA LOBO null (rosangelalobo@btu.unesp.br) on 2020-04-17T21:46:04Z (GMT) No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Made available in DSpace on 2020-04-17T21:46:04Z (GMT). No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Previous issue date: 2020-02-20 Made available in DSpace on 2020-04-17T21:46:04Z (GMT). No. of bitstreams: 1 garcia_nap_me_bot_par.pdf: 681433 bytes, checksum: c850e537746498af4069016d52f208d9 (MD5) Previous issue date: 2020-02-20 Introdução: Considerando que no Brasil, observou-se um crescimento exponencial das demandas judiciais relacionadas aos serviços prestados pelos profissionais da saúde, entende-se necessário um estudo aprofundado à respeito do erro médico com abordagem direta a esse profissional, carecedor de tratamento especial e protetivo sempre que, diante das falhas oriundas do seu exercício profissional, forem verificados fatores de imprevisibilidade capazes de comprometer a exitosa prestação do serviço ofertado. Necessária também, a abordagem no presente trabalho, das mudanças na relação entre o profissional da saúde e o paciente, as prerrogativas de facilitação do acesso ao judiciário, o atendimento aos protocolos clínicos preestabelecidos, dentre outras variáveis, sendo que tais pontos são relevantes para o crescente aumento das ações judiciais e, por consequência, a forma que as decisões são proferidas ao apreciar problemáticas dessa natureza. Objetivos: Descrever acerca da problemática do erro médico quando analisado sob a ótica da responsabilidade civil. Explorar a teoria da responsabilidade civil, a extensão das variáveis de responsabilidade, legislações pertinentes, além de comparar decisões proferidas no Brasil e em outros países. Metodologia: Fora realizada uma revisão da literatura existente sobre o tema e assim elaborada uma dissertação que reuniu e analisou doutrinas acerca do erro médico. Dentre os materiais que foram utilizados, estão as legislações nacionais e internacionais abrangentes ao tema e estudos jurídicos existentes que foram obtidos por meio de livros, Códigos, Constituição da República, artigos publicados em revistas especializadas, Acórdãos de Tribunais Superiores e textos publicados na Internet. A pesquisa se desenvolveu da seguinte forma: Levantamento bibliográfico inerente ao tema; Estudo crítico do material doutrinário levantado; Obtenção e análise da legislação nacional e internacional pertinente; Análise de textos alternativos referente ao tema em questão. Resultados: Após verificados diversos estudos, temos que uma série de fatores pode influenciar na ocorrência do “erro médico”, podendo ocorrer deficiência da formação profissional, falha na relação profissional-paciente e a falta de condições ideais de trabalho. Temos assim, que se torna necessário priorizar o ensino no sentido de fortalecer virtudes que favorecem uma melhora na relação com o paciente, como humanismo e empatia, palavras primordiais para uma possível melhora deste quadro em nosso país. Conclusão: Se mostra muito importante o estudo e a conscientização da problemática que envolve uma possível indústria do “erro médico” no Brasil. Temos que imaginar que as consequências da fomentação da proteção do paciente a qualquer custo trará herança desastrosa para a comunidade. Portanto, indispensável que os operadores do direito tenham maior interação sobre o assunto, extinguindo pré-suposições e pré-conceitos, amparando esta classe nas demandas infundadas de erros imaginários perpetradas por pacientes. Um suposto erro médico pode ser uma consequência de várias situações, mas como visto, não temos dúvidas de que estabelecer uma boa relação entre profissional e paciente, baseada no respeito pessoa, poderia evitar suas referidas ocorrências e, por consequência, a maioria dos processos judiciais. Introduction: Considering that in Brazil, there has been an exponential increase in the legal demands related to the services provided by health professionals, it is necessary to conduct an in-depth study about medical error with a direct approach to this professional, who needs special and protective treatment, whenever, in the face of failures arising from their professional practice, unpredictable factors are verified that can compromise the successful provision of the service offered. It is also necessary, the approach in the present work, of the changes in the relationship between the health professional and the patient, the prerogatives of facilitating access to the judiciary, the attendance to the pre-established clinical protocols, among other variables, and such points are relevant to the increasing increase in lawsuits and, consequently, the way in which decisions are rendered when considering problems of this nature. Objectives: To describe the problem of medical error when analyzed from the perspective of civil liability. Explore the theory of civil liability, the extent of liability variables, relevant legislation, and compare decisions made in Brazil and other countries. Methodology: A review of the existing literature on the topic had been carried out and a dissertation was prepared, which brought together and analyzed doctrines about medical error. Among the materials that were used, there are national and international laws covering the theme and existing legal studies that were obtained through books, Codes, the Constitution of the Republic, articles published in specialized magazines, Judgments of Superior Courts and texts published on the Internet. The research was developed as follows: Bibliographic survey inherent to the theme; Critical study of the doctrinal material raised; Obtaining and analyzing relevant national and international legislation; Analysis of alternative texts related to the topic in question. Results: After verifying several studies, we have that a series of factors can influence the occurrence of the “medical error”, with the possibility of professional training deficiency, failure in the professional-patient relationship and the lack of ideal working conditions. We have, therefore, that it is necessary to prioritize teaching in order to strengthen virtues that favor an improvement in the relationship with the patient, such as humanism and empathy, essential words for a possible improvement of this situation in our country. Conclusion: The study and awareness of the problem surrounding a possible “medical error” industry in Brazil is very important. We have to imagine that the consequences of promoting patient protection at any cost will bring disastrous inheritance to the community. Therefore, it is essential that legal operators have greater interaction on the subject, extinguishing pre-assumptions and preconceptions, supporting this class in the unfounded demands of imaginary errors perpetrated by patients. An alleged medical error can be a consequence of several situations, but as seen, we have no doubt that establishing a good relationship between professional and patient, based on respect for the person, could prevent their occurrences and, consequently, most processes judicial.

Using the AdaBoost method may increase the accuracy and reliability of a framework for daily activities and environment recognition. Mobile devices have several types of sensors, including motion, magnetic, and location sensors, that allow accurate identification of daily activities and environment. This paper focuses on the review of the studies that use the AdaBoost method with the sensors available in mobile devices. This research identified the research works written in English about the recognition of daily activities and environment recognition using the AdaBoost method with the data obtained from the sensors available in mobile devices that were published between 2012 and 2018. Thus, 13 studies were selected and analysed from 151 identified records in the searched databases. The results proved the reliability of the method for daily activities and environment recognition, highlighting the use of several features, including the mean, standard deviation, pitch, roll, azimuth, and median absolute deviation of the signal of motion sensors, and the mean of the signal of magnetic sensors. When reported, the analysed studies presented an accuracy higher than 80% in recognition of daily activities and environments with the Adaboost method.

Submitted by Nuno Garcia (nmgs@ubi.pt) on 2020-01-24T13:57:44Z No. of bitstreams: 1 2020 - Activities of Daily Living and Environment Recognition Using Mobile Devices A Comparative Study.pdf: 261493 bytes, checksum: d0d7bd634352d3d47cf054db90d7e5bd (MD5) Approved for entry into archive by Pessoa (pfep@ubi.pt) on 2020-01-24T15:40:43Z (GMT) No. of bitstreams: 1 2020 - Activities of Daily Living and Environment Recognition Using Mobile Devices A Comparative Study.pdf: 261493 bytes, checksum: d0d7bd634352d3d47cf054db90d7e5bd (MD5) Approved for entry into archive by Pessoa (pfep@ubi.pt) on 2020-01-24T15:45:23Z (GMT) No. of bitstreams: 1 2020 - Activities of Daily Living and Environment Recognition Using Mobile Devices A Comparative Study.pdf: 261493 bytes, checksum: d0d7bd634352d3d47cf054db90d7e5bd (MD5) Made available in DSpace on 2020-01-24T15:45:23Z (GMT). No. of bitstreams: 1 2020 - Activities of Daily Living and Environment Recognition Using Mobile Devices A Comparative Study.pdf: 261493 bytes, checksum: d0d7bd634352d3d47cf054db90d7e5bd (MD5) Previous issue date: 2020 info:eu-repo/semantics/publishedVersion