Instance Learning Research Articles

Automatic Generation Method for Training Dataset of Instance Segmentation for Irregularly Shaped Object Picking

In the field of robotics, robotic bin picking has been studied extensively. In this paper, we consider an irregularly shaped object as a grasping target. For detecting irregularly shaped objects, instance segmentation based on deep learning is one of the promising methods. One of the challenges for deep learning-based methods is how to reduce the time and effort to prepare the dataset for training. In this paper, we propose a method to automatically generate a dataset for learning instance segmentation using only information available from public image databases. The proposed method achieves mean average precision (mAP) of 0.85 for the automatically generated test data. It also showed mAP of 0.65 for the test data generated using untrained irregularly shaped objects, and achieved a success rate of more than 98% in picking experiments with the robot.

Segmentation of MRI Brain Tumor Image using Optimization based Deep Convolutional Neural networks (DCNN)

Abstract Segmentation of brain image should be done accurately as it can help to predict deadly brain tumor disease so that it can be possible to control the malicious segments of brain image if known beforehand. The accuracy of the brain tumor analysis can be enhanced through the brain tumor segmentation procedure. Earlier DCNN models do not consider the weights as of learning instances which may decrease accuracy levels of the segmentation procedure. Considering the above point, we have suggested a framework for optimizing the network parameters such as weight and bias vector of DCNN models using swarm intelligent based algorithms like Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Gray Wolf Optimization (GWO) and Whale Optimization Algorithm (WOA). The simulation results reveals that the WOA optimized DCNN segmentation model is outperformed than other three optimization based DCNN models i.e., GA-DCNN, PSO-DCNN, GWO-DCNN.

Analytical study on COVID-19 to predict future infected cases ratio in India using Machine Leaning

COVID-19 is real a worldwide terrific problem. This paper focuses on the different aspects of data analytics and visualization by using various datasets supported by authorized sources. It also discusses the practical aspects using open source tools and python library support. Here chapter focuses on comparative analysis also. It also visualize analytical aspects by different aspects such as country wise, date wise and so on. In this paper, the COVID infected cases and its reaction on people will be discussed. This case study will predict the COVID-19 infected cases and death ratio with symptoms in future. This paper focus on data visualization, data analytics and comparative study based on practical aspects. Machine Learning plays a vital role to predict the cases by providing learning instances.

Instance Segmentation for Large, Multi-Channel Remote Sensing Imagery Using Mask-RCNN and a Mosaicking Approach

Instance segmentation is the state-of-the-art in object detection, and there are numerous applications in remote sensing data where these algorithms can produce significant results. Nevertheless, one of the main problems is that most algorithms use Red, Green, and Blue (RGB) images, whereas Satellite images often present more channels that can be crucial to improve performance. Therefore, the present work brings three contributions: (a) conversion system from ground truth polygon data into the Creating Common Object in Context (COCO) annotation format; (b) Detectron2 software source code adaptation and application on multi-channel imagery; and (c) large scene image mosaicking. We applied the procedure in a Center Pivot Irrigation System (CPIS) dataset with ground truth produced by the Brazilian National Water Agency (ANA) and Landsat-8 Operational Land Imager (OLI) imagery (7 channels with 30-m resolution). Center pivots are a modern irrigation system technique with massive growth potential in Brazil and other world areas. The round shapes with different textures, colors, and spectral behaviors make it appropriate to use Deep Learning instance segmentation. We trained the model using 512 × 512-pixel sized patches using seven different backbone structures (ResNet50- Feature Pyramid Network (FPN), Resnet50-DC5, ResNet50-C4, Resnet101-FPN, Resnet101-DC5, ResNet101-FPN, and ResNeXt101-FPN). The model evaluation used standard COCO metrics (Average Precision (AP), AP50, AP75, APsmall, APmedium, and AR100). ResNeXt101-FPN had the best results, with a 3% advantage over the second-best model (ResNet101-FPN). We also compared the ResNeXt101-FPN model in the seven-channel and RGB imagery, where the multi-channel model had a 3% advantage, demonstrating great improvement using a larger number of channels. This research is also the first with a mosaicking algorithm using instance segmentation models, where we tested in a 1536 × 1536-pixel image using a non-max suppression sorted by area method. The proposed methodology is innovative and suitable for many other remote sensing problems and medical imagery that often present more channels.

Three-dimensional photogrammetry with deep learning instance segmentation to extract berry fruit harvestability traits

Fruit cluster characteristics such as compactness, maturity, berry number, and berry size, are important phenotypic traits associated with harvestability and yield of blueberry genotypes and can be used to monitor berry development and improve crop management. The goal of this study was to develop a complete framework of 3D segmentation for individual blueberries as they develop in clusters and to extract blueberry cluster traits. To achieve this goal, an image-capturing system was developed to capture blueberry images to facilitate 3D reconstruction and a 2D-3D projection-based photogrammetric pipeline was proposed to extract berry cluster traits. The reconstruction was performed for four southern highbush blueberry cultivars (‘Emerald’, ‘Farthing’, ‘Meadowlark’ and ‘Star’) with 10 cluster samples for each cultivar based on photogrammetry. A minimum bounding box was created to surround a 3D blueberry cluster to calculate compactness as the ratio of berry volume and minimum bounding box volume. Mask R-CNN was used to segment individual blueberries with the maturity property from 2D images and the instance masks were projected onto 3D point clouds to establish 2D-3D correspondences. The developed trait extraction algorithm was used to segment individual 3D blueberries to obtain berry number, individual berry volume, and berry maturity. Berry maturity was used to calculate cluster maturity as the ratio of the mature berry (blue colored fruit) number and the total berry (blue, reddish, and green colored fruit) number comprising the cluster. The accuracy of determining the fruit number in a cluster is 97.3%. The linear regression for cluster maturity has a R2 of 0.908 with a RMSE of 0.068. The cluster berry volume has a RMSE of 2.92 cm3 compared with the ground truth, indicating that the individual berry volume has an error of less than 0.292 cm3 for clusters with a berry number greater than 10. The statistical analyses of the traits for the four cultivars reveals that, in the middle of April, ‘Emerald’ and ‘Farthing’ were more compact than ‘Meadowlark’ and ‘Star’, and the mature berry volume of ‘Farthing’ was greater than ‘Emerald’ and ‘Meadowlark’, while ‘Star’ had the smallest mature berry size. This study develops an effective method based on 3D photogrammetry and 2D instance segmentation that can determine blueberry cluster traits accurately from a large number of samples and can be used for fruit development monitoring, yield estimation, and harvest time prediction.

Connections

The human brain contains around 86 billion nerve cells and about as many glial cells [1]. In addition, there are about 100 trillion connections between the nerve cells alone. While mapping all the connections of a human brain remains out of reach, scientists have started to address the problem on a smaller scale. The term artificial neural networks (ANNs or simply neural networks (NNs), encompassing a family of nonlinear computational methods that, at least in the early stage of their development, were inspired by the functioning of the human brain. Indeed, the first ANNs were nothing more than integrated circuits devised to reproduce and understand the transmission of nerve stimuli and signals in the human central nervous system [2]. The correct way of doing it is to the first study human behavior. The human brain has a biological neural network that has billions of interconnections. As the brain learns, these connections are either formed, changed or removed, similar to how an artificial neural network adjusts its weights to account for a new training example. This complexity is the reason why it is said that practice makes one perfect since a greater number of learning instances allow the biological neural network to become better at whatever it is doing. Depending upon the stimulus, only a certain subset of neurons are activated in the nervous system. Recently, Moreau et al., [3] published an interesting paper studying how artificial intelligence can help doctors and patients with meningiomas make better treatment decisions, according to a new study. They demonstrated that their models were capable of predicting meaningful individual-specific clinical outcome variables and show good generalizability across the Surveillance, Epidemiology, and End Results (SEER) database to predict meningioma malignancy and survival after specific treatments. Statistical learning models were trained and validated on 62,844 patients from the SEER database and a model scoring for the malignancy model was performed using a series of metrics. A free smartphone and web application were also provided for readers to access and test the predictive models (www.meningioma.app). The use of artificial intelligence techniques is gradually bringing efficient theoretical solutions to a large number of real-world clinical problems related to the brain (4). Specifically, recently, thanks to the accumulation of relevant data and the development of increasingly effective algorithms, it has been possible to significantly increase the understanding of complex brain mechanisms. The researchers' efforts are creating increasingly sophisticated and interpretable algorithms, which could favor a more intensive use of “intelligent” technologies in practical clinical contexts. Brain and machine working together will improve the power of these methods to make individual-patient predictions could lead to improved diagnosis, patient counseling, and outcomes.

Pedagogical Knowledge: A Reflection on the Pedagogical Praxis of Teachers during the Initial Phase of Their Profession

The present article aims to understand and interpret the experiences that new teachers recall and learn from their initial teacher training. This will be conducted by understanding that knowledge is built daily through the exercise of teaching. In such a way that the experience and wisdom of pedagogical knowledge are consolidated through action and, given this context, praxis is not just an object of reflection, but an instance of learning for the exercise of teaching. Regarding methodology, this study stands from the interpretive paradigm and qualitative approach. To collect evidence, eight focus groups and twelve in-depth interviews were conducted with teachers who graduated from a Chilean university located in the center-south of the country. The sample was composed of a total of 53 teachers. As for data analysis, certain notions raised by grounded theory were utilized. The results of the study reveal the necessity—manifested by teachers—regarding initial training that allows for the construction of new knowledge and spaces to increase critical-reflective thinking. On the other hand, given their current pedagogical exercise within the classroom, teachers consider that initial teacher training must improve its theoretical-practical relationship, modify the curriculum and evaluation processes that are developed at university, consistently establish the university-school link.

Liking as far as you like yourself: Exploring the Self-Referencing effect across multiple intersecting regularities and its relationship with self-esteem

Classic instances of evaluative learning require the spatio-temporal contingency between source and target stimuli. However, people can learn to like stimuli in a more indirect fashion. Moreover, many of our preferences are self-referential: we tend to like the objects that are related to ourselves. For instance, it is demonstrated that performing the Self-Referencing task, a categorization task based on intersecting regularities between the self and a target stimulus, results in a transfer of positivity from the self to the target. Unexplored so far is the extent to which intersecting regularities can be exploited to form even less direct connections between the self and the target, which ultimately results in a preference for the latter. In two experiments, we tested the indirect SR effect. To this aim, we set up a two-step learning procedure in which, after completion of the SR task, participants underwent an additional task, where target stimuli categorized in the SR shared a behavioral response with novel targets. Experiment 1 (N = 241) showed significant direct and indirect SR effects, both implicitly and explicitly, irrespectively of the type of targeted stimuli (i.e., groups or brands). Experiment 2 (N = 174) replicated such findings and showed that explicit, but not implicit, self-esteem moderated the SR effect. Although at a descriptive level the moderating role of explicit self-esteem showed stronger on the indirect SR effect, no statistical evidence was found to support a differential impact on either direct or indirect SR. We discuss the power of IR to transfer self-positivity indirectly across stimuli, the role of self-esteem, and the theoretical implications of our findings.

EU economic governance and Covid-19: policy learning and windows of opportunity

ABSTRACT This article analyses the extent to which the Covid-19 pandemic crisis represents a window of opportunity towards fundamental change in the economic governance of the European Union (EU). Adopting a historical institutionalist (HI) perspective and drawing insights from the policy learning literature, we argue that contingent learning immediately took place and policy entrepreneurs took important decisions recognising the new crisis as an existential threat for the EU. Further, the pandemic crisis support fund and the ECB pandemic emergency purchase programme represent instances of single loop learning that leave the fundamentals of economic governance untouched. However, and in contrast to the Euro area crisis response, the adoption of the Recovery and Resilience Facility (RRF) represents a bold decision and suggests double-loop learning. It is argued that the Covid-19 crisis is a critical juncture for the EU. As a result, EU economic governance ceases to be limited to its regulatory function and is now complemented by a redistributive function as well.

“Good Robot!”: Efficient Reinforcement Learning for Multi-Step Visual Tasks with Sim to Real Transfer

Current Reinforcement Learning (RL) algorithms struggle with long-horizon tasks where time can be wasted exploring dead ends and task progress may be easily reversed. We develop the SPOT framework, which explores within action safety zones, learns about unsafe regions without exploring them, and prioritizes experiences that reverse earlier progress to learn with remarkable efficiency. The SPOT framework successfully completes simulated trials of a variety of tasks, improving a baseline trial success rate from 13% to 100% when stacking 4 cubes, from 13% to 99% when creating rows of 4 cubes, and from 84% to 95% when clearing toys arranged in adversarial patterns. Efficiency with respect to actions per trial typically improves by 30% or more, while training takes just 1-20k actions, depending on the task. Furthermore, we demonstrate direct sim to real transfer. We are able to create real stacks in 100% of trials with 61% efficiency and real rows in 100% of trials with 59% efficiency by directly loading the simulation-trained model on the real robot with no additional real-world fine-tuning. To our knowledge, this is the first instance of reinforcement learning with successful sim to real transfer applied to long term multi-step tasks such as block-stacking and row-making with consideration of progress reversal. Code is available at https://github.com/jhu-lcsr/good_robot .

A Quadruplet Loss for Enforcing Semantically Coherent Embeddings in Multi-Output Classification Problems

This article describes one objective function for learning semantically coherent feature embeddings in multi-output classification problems, i.e., when the response variables have dimension higher than one. Such coherent embeddings can be used simultaneously for different tasks, such as identity retrieval and soft biometrics labelling. We propose a generalization of the triplet loss that: 1) defines a metric that considers the number of agreeing labels between pairs of elements; 2) introduces the concept of similar classes, according to the values provided by the metric; and 3) disregards the notion of anchor, sampling four arbitrary elements at each time, from where two pairs are defined. The distances between elements in each pair are imposed according to their semantic similarity (i.e., the number of agreeing labels). Likewise the triplet loss, our proposal also privileges small distances between positive pairs. However, the key novelty is to additionally enforce that the distance between elements of any other pair corresponds inversely to their semantic similarity. The proposed loss yields embeddings with a strong correspondence between the classes centroids and their semantic descriptions. In practice, it is a natural choice to jointly infer coarse (soft biometrics) + fine (ID) labels, using simple rules such as k-neighbours. Also, in opposition to its triplet counterpart, the proposed loss appears to be agnostic with regard to demanding criteria for mining learning instances (such as the semi-hard pairs). Our experiments were carried out in five different datasets (BIODI, LFW, IJB-A, Megaface and PETA) and validate our assumptions, showing results that are comparable to the state-of-the-art in both the identity retrieval and soft biometrics labelling tasks.

AUTOMATED MARINE OIL SPILL DETECTION USING DEEP LEARNING INSTANCE SEGMENTATION MODEL

Abstract. This study developed a novel deep learning oil spill instance segmentation model using Mask-Region-based Convolutional Neural Network (Mask R-CNN) model which is a state-of-the-art computer vision model. A total of 2882 imageries containing oil spill, look-alike, ship, and land area after conducting different pre-processing activities were acquired. These images were subsequently sub-divided into 88% training and 12% for testing, equating to 2530 and 352 images respectively. The model training was conducted using transfer learning on a pre-trained ResNet 101 with COCO data as a backbone in combination with Feature Pyramid Network (FPN) architecture for the extraction of features at 30 epochs with 0.001 learning rate. The model’s performance was evaluated using precision, recall, and F1-measure which shows a higher performance than other existing models with value of 0.964, 0.969 and 0.968 respectively. As a specialized task, the study concluded that the developed deep learning instance segmentation model (Mask R-CNN) performs better than conventional machine learning models and semantic segmentation deep learning models in detection and segmentation of marine oil spill.

Pandemia por COVID-19 y la virtualización de las aulas: La importancia del juego

Actualmente, la pandemia que afecta a la totalidad del planeta ha producido cambios. Poniéndonos en contexto y según Arrizabalaga (1992): el vocablo “pandemia” etimológicamente procede de la expresión griega pandêmonnosêma, traducida como enfermedad del pueblo entero, ésta ha hecho tomar diferentes decisiones para el abordaje del proceso enseñanza-aprendizaje. Este ensayo busca cotejar la importancia del juego y de la generación de instancias de aprendizaje lúdicas en esta nueva metodología impuesta por la contingencia sanitaria. Para lograr el análisis del objeto de estudio se realizó una revisión sistemática desde fuentes formales, de los efectos del uso de las pantallas, de la digitalización docente y de cómo potenciar estos instrumentos con el fin del logro de los resultados de aprendizajes.

Approach to the hidden curriculum through narrative medicine in the training of medical students

Hidden curricula are those contents learned that do not belong to the formal curriculum. Narrative medicine allows knowledge of health professionals' experiences. Approaching to the hidden curriculum can allow modifying or reinforcing instances of the nonformal learning. Analyze narrative medical students' productions to identify hidden curriculum contents. A cross-sectional study was conducted with Pediatric students during 2017. All the students of the course (14) participated and performed 50 narrative productions: 53 % corresponded to the hospitalization context, 17 % to ambulatory care, 17 % to the emergency center and 13 % to neonatology. The main topics described were doctor-patient communication, importance of family inclusion in the pediatric consultation, empathy, impact of chronic disease in children and communication of bad news. Narrative productions allowed an approximation to the hidden curriculum through reflections on professional actions.

A novel deep learning instance segmentation model for automated marine oil spill detection

The visual similarity of oil slick and other elements, known as look-alike, affects the reliability of synthetic aperture radar (SAR) images for marine oil spill detection. So far, detection and discrimination of oil spill and look-alike are still limited to the use of traditional machine learning algorithms and semantic segmentation deep learning models with limited accuracy. Thus, this study developed a novel deep learning oil spill detection model using computer vision instance segmentation Mask-Region-based Convolutional Neural Network (Mask R-CNN) model. The model training was conducted using transfer learning on the ResNet 101 on COCO as backbone in combination with Feature Pyramid Network (FPN) architecture for feature extraction at 30 epochs with 0.001 learning rate. Testing of the model was conducted using the least training and validation loss value on the withheld testing images. The model’s performance was evaluated using precision, recall, specificity, IoU, F1-measure and overall accuracy values. Ship detection and segmentation had the highest performance with overall accuracy of 98.3%. The model equally showed a higher accuracy for oil spill and look-alike detection and segmentation although oil spill detection outperformed look-alike with overall accuracy values of 96.6% and 91.0% respectively. The study concluded that the deep learning instance segmentation model performs better than conventional machine learning models and deep learning semantic segmentation models in detection and segmentation.

Can the stimulus processing assumptions of the sometimes-opponent-process (SOP) model explain instances of contextual learning?

One of the most persisting assertions in Allan Wagner's view of conditioning is that the environment or context in which significant events occur can develop an association with these events, more or less in the same way as conditioned and unconditioned stimuli become associated with each other. He was drawn to this idea by evidence of contextual fear conditioning, contingency effects, some instances of context-specificity of long-term habituation, and latent inhibition. From a theoretical point of view, however, homologizing contexts to conditioned stimuli is not as simple as it seems, especially when quantitative theories are involved, as is the case of Wagner's work. It might be, for instance, that contexts cannot be represented merely as long-duration conditioned stimuli, in which case, no net contextual learning can occur due to the context being less correlated with reinforcement than with nonreinforcement. In this article, we use Wagner's sometimes-opponent-process model of conditioning to comment on the quantitative nature of this challenge. Also, based on an idea sketched by Mazur and Wagner, we describe a set of quantitative strategies that might be usefully considered to solve this dilemma within the general framework of Wagner's theory. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

La regulación del cannabis medicinal en Uruguay y los desafíos para la conformación de circuitos innovativos

A partir de la promulgación de la Ley Nº 19.172 Uruguay comenzó un proceso nacional de regulación de cannabis. Ello involucró modificaciones institucionales trascendentes como la creación del Instituto de Regulación y Control del Cannabis (IRCCA). Así también, se sucedieron instancias de aprendizaje, adaptación y momentos de incertidumbre política y sucesivos bloqueos provenientes de diversos sectores. El artículo busca analizar cómo el esquema elegido para la regulación del mercado de cannabis genera incentivos para la instalación de circuitos innovativos (CI) y, concomitantemente, micro-procesos innovadores destinados a resolver problemas de distinto tipo. Partiendo de una aproximación metodológica cualitativa que comprendió, por un lado, la realización de entrevistas a informantes calificados y protagonistas de la implementación de la política, así como actores de la gobernanza de la política y por otro la revisión de fuentes documentales, el documento argumenta que el diseño institucional y regulatorio que se desprende del marco normativo escogido habrían generado una serie de nudos o complejidades que condicionaron la aparición de CI.

Aprender o ofício acadêmico: impressões etnográficas sobre espaços de formação profissional no Ensino Superior

Neste ensaio etnográfico discorremos quanto ao processo complexo de aprendizagem do ofício acadêmico, a partir de uma exemplificação com uma observação etnográfica, em que operou-se o exercício de estranhar o familiar para compreender e refletir sobre os modos como ocorrem os processos de interação, socialização e aprendizagem da profissionalidade docente no interior da instituição acadêmica com suas rotinas, eventos, processualidades, ritualidades etc., entendendo-a como uma comunidade de prática. Observar o familiar e refletir sobre nossos próprios processos de inserção acadêmica e profissional nos permite perceber que há instâncias de aprendizado que não se restringem ao espaço da sala de aula, embora este mesmo se configure como espaço primordial na formação docente, e, assim, valorizar e dar importância aos inúmeros eventos e atividades acadêmicas que compõem o cenário formativo dos professores, considerando que as “trocas” são reconhecidas como aprendizagem situada periférica a qualquer campo profissional.

Replay-peak attention chart: A performance measure for improving online video lecture design

The ability to replay selected video segments is a major advantage of online video lectures. Replay is a learning instance that reflects active engagement. This paper develops the ‘replay-peak attention chart’ as a new performance measure of learner’s attention, based on the control chart concept, which is used for Statistical Process Control (SPC) in operations management. This study follows the design science research paradigm and employs a mixed methods methodology, combining quantitative learning analytics with qualitative analysis of notable segment replay instances by viewers of online video lectures. An analysis of a successful Massive Open Online Course (MOOC), titled “Negotiation Management” provides a proof-of-concept for the replay-peak attention chart, as a visual heuristic tool for identifying notable learning instances. The MOOC includes Educational Entertainment (edutainment) in the form of negotiation simulations which are presented as sitcoms, and are meant to increase learner engagement. From an attention economy perspective, the replay-peak attention chart may help instructors and designers to focus their limited attention resources on segments of online video lecture sessions that may require pedagogical interventions. This paper critically discusses the replay-peak attention chart conceptualization and its initial proof-of-concept. It suggests future research directions for substantiating the replay-peak attention chart, and investigating the effect of edutainment on online learning. The replay-peak attention chart is a dynamic descriptive performance measure, which has a promising potential to improve the design of effective online video lectures as an e-learning resource.

Concurrent service access and management framework for user-centric future internet of things in smart cities

Future Internet of Things (FIoT) is a service concentric distributed architecture that is used by the smart city users for information sharing and access. The design of FIoT focuses in achieving reliable service and response to the growing user demands through different interoperability features. In this manuscript, concurrent service access and management framework is introduced to improve the swiftness in user concentric request processing. Based on the availability of the services and the density of the users, the concurrency in information access is provided to the users in a reliable manner. The framework incorporates convolution neural learning process in linear and differential manner for improving the access and service usage rates of the requesting users. The access sessions are differentiated for the accessible and offloaded requests to the available service providers based on the learning instances. The proposed framework is assessed using the metrics access rate, service usage rate, access delay, time lag, and failure ratio.