Brazil

Glaucoma is a group of eye diseases characterized by progressive degeneration of the optic nerve, which can lead the patient to blindness, however it can be prevented through diagnosis and treatment of the disease. It is the leading cause of irreversible blindness worldwide and is expected to affect 76 million by 2020. The cup-to-disc ratio technique is a structural examination of the optic nerve used to glaucoma diagnose, and requires an accurate segmentation of the optic disc and cup. Thus, this work proposes an automated method for cup segmentation in fundus images of DRISHTI-GS dataset using a combination of the green channel of RGB image and the provided optic disc mask as input to a modified U-Net convolutional neural network. The results achieved were promising, yielding a mean dice coefficient of 94%, much better than state-of-the-art algorithms. The method could be used for glaucoma diagnosis.

IWSSIP 2020

Mask Overlaying: A Deep Learning Approach for Individual Optic Cup Segmentation from Fundus Image

drishti-gs

optic cup

glaucoma

u-net

segmentation

deep learning


IWSSIP is an international conference that brings together researchers and developers from both academia and industry reporting the latest scientific and theoretical advances to discuss and debate major issues and state-of-the-art aspects on systems, signals and image processing.

###### For the session discussion, each author will be seated at individual breakout tables.  Please find which table they will be at here:
###### [QA table assignments in Gather.town](https://drive.google.com/file/d/1hBHOD_KF7NSsaTcpYcHCZslXJvQVxJQe/view?usp=sharing)

IWSSIP 2021

technical paper

The 27th International Conference on Systems, Signals and Image Processing.

In this presentation, the comparison between deep learning methods and feature extraction algorithms is presented. The principle of Grey-Level Co-occurrence Matrix (GLCM) and its modifications are used for our research. The main idea was to design a method for the description of combined features and textures. The texture classification process is carried out with the robust support vector machine classifier (SVM). We compare these feature extraction methods with proposed Convolutional Neural Networks (CNN). This proposed network contains 25 layers. Finally, the all evaluation and comparison of color texture retrieval results for all used methods are presented. The all feature extraction algorithms and proposed CNN have been tested on two different color texture datasets (Outex and Vistex datasets).

The Comparison of Color Texture Features Extraction based on 1D GLCM with Deep Learning Methods

Efficient management of cloud traffic leads to resource sharing among clients. To meet clients’ performance restrictions, however, cloud providers adopt strict resource allocation that implies idle resources as clients’ bandwidth thresholds are over-provisioned. We propose a dynamic mechanism to provide resources on-demand in a multitenant data center. The mechanism relies on Q-learning multi-agent for managing each client access to the cloud resources. The proposed mechanism either measure the throughput continuously or map the CPU usage into bandwidth usage. We assess our mechanism in an emulated environment, in which a network controller provisions bandwidth. Results show that the mechanism reduces cloud idleness, allowing low-priority clients to use bandwidth while there is idle capacity. Our mechanism incurs low processing overhead, even when the number of states and action space grows, while the memory cost per agent increases, peaking at 300 kB for 200 states and actions. When mapping CPU usage into bandwidth usage, the mechanism achieves fair bandwidth sharing at the cost of a small increase on the convergence time to an optimal policy.

A Lightweight Reinforcement-Learning-based Mechanism for Bandwidth Provisioning on Multitenant Data Center

Wind energy is promoting the disruption of the current energy model, as the world begins to reduce the use of non-renewable energy sources. At the core of this energetic revolution, the wind turbines are designed to capture the wind's kinetic energy to produce electricity. Wind turbines are complicated and expensive systems whose operating signals are converted into high dimensional data for health management. This signal-to-data conversion involves a variety of constraints that produces an imperfect dataset as an output.
In recent years, many semi-supervised learning approaches were developed to detect faults in a wind turbine preemptively. In this paper, we propose a workflow to detect faults based on two hot research topics in data science: (a) automatic feature selection; and (b) semi-supervised classification for fault detection. The multi-objective optimization framework Non-dominated Sorting Genetic Algorithm II (NSGA-II) was adopted to select a subset of features and hyperparameters to perform the semi-supervised fault detection using both a Soft-Label and a Binary support vector machine (SVM). Our experiments were carried out using a dataset containing unlabelled records of 5 2.0MW wind turbines. We discuss the results from NSGA-II based on the trade-off between the number of selected features and the detection performance. We also select the algorithm configuration that achieved the best savings by predicting a faulty system before its collapse.

Wind Turbine Fault Detection: A Semi-Supervised Learning Approach With Automatic Evolutionary Feature Selection

In order to minimize evasion losses, socioeconomic policies and programs were created to assist and support actions seeking to maximize the number of students that graduate in suitable program time. This work presents a methodology that aims to predict evasion by using machine learning. Our approach was able to classify student abandonment with an average f-score and accuracy results above 95%. Our approach combines a decision tree alongside a genetic algorithm and cluster stratified sampling.

EvolveDTree: Analyzing Student Dropout in Universities

This paper presents the design and implementation details of a system build-up by using off-the-shelf algorithms for urban video analytics. The system allows the connection to public video surveillance camera networks to obtain the necessary information to generate statistics from urban scenarios (e.g., amount of vehicles, type of cars, direction, numbers of persons, etc.). The obtained information could be used not only for traffic management but also to estimate the carbon footprint of urban scenarios. As a case study, a university campus is selected to evaluate the performance of the proposed system. The system is implemented in a modular way so that it is being used as a testbed to evaluate different algorithms. Implementation results are provided showing the validity and utility of the proposed approach.

Off-the-Shelf Based System for Urban Environment Video Analytics

Dynamic smart cities models should provide enhanced descriptions on the corresponding systems, processes and services to facilitate their continuous improvement, learning and innovation, oriented to satisfy the increasing necessities of the involved educated citizens. This paper proposes a cognitive enterprise architecture framework (EAF) for smart cities, supported by a hybrid model based on a geographic information system (GIS) and associated graph-database (GDB). The corresponding framework allows managing and integrating: spatial and nonspatial data, dynamic and static data, numeric and semantic data, physical and conceptual data, information and knowledge, and associated relations of the corresponding city systems, environment and context. The augmented space model considered is based on augmented reality (AR) and augmented space (AS) principles, which includes augmented systems, channels, images, models and maps. This cognitive space can improve the governance, observability and controllability, interactions, visibility, transparency, monitoring, and review of the corresponding city services. The model supports the selected enterprise architecture framework for continuously aligning the diverse city stakeholder goals, technologies, ecosystems and roadmaps. Several application factors, at different levels, are analyzed to demonstrate the capabilities and possibilities of the proposed model.

The Augmented Space of a Smart City

Smart parking systems can manage real-time city resources, and geo-locally provides them to drivers. It involves on– and off–street parking management; that is why smart parking has been considered promising inception and a leading paradigm of a smart city. The design of a smart city ecosystem, especially on urban mobility, has a tremendous impact on resource status and city capacities. Although its high capacity to work for quality of experience and Business Intelligence, works about smart parking in the literature do not take advantage of this in its proposes. This work proposes the BEM framework that intends to take advantage of the car park Management and improve Quality Experience and Business Intelligence being an essential vector for the advancement of smart cities. A proof of concept was implemented, proving the solution viability.

BEM: A Framework based on Business Intelligence, Quality of Experience, and Car Park Management

Evaluation of Output Representations in Neural Network-based Trajectory Predictions Systems

Intelligent transport support systems have had a major impact on people’s urban mobility. In large urban centers, transportation services still need ways to optimize vehicle supply in certain areas, according to the demand in each of them. This work proposes a stream learning methodology for taxi demand prediction, that tackles concept drift detection on a real taxi data stream.

A Methodology for Taxi Demand Prediction Using Stream Learning

Smart cities should be built on harmonious relationships with the environment, with rational measures to use environmental resources in many ways. Resources essential to human life such as water and energy must have special attention. With the advance of technologies use, human being becomes one of the main promoters of information and solutions for smart cities. Stream machine learning is an approach where data becomes available sequentially, called data stream, which adapts the prediction model with fixed window-sized data stream. The objective of this work is to analyze the effectiveness of buildings energy consumption using stream learning. To achieve this goal, MOA tool was used to evaluate the performance of the selected stream learning algorithms on an an open database. Batch algorithms were also used as testbed, to analyze the performance of stream algorithms related to them

Premium content

Downloads

Next from IWSSIP 2020

The Comparison of Color Texture Features Extraction based on 1D GLCM with Deep Learning Methods

Similar lecture

Segmentation of signiﬁcant regions in retinal images: perspective of U-Net network through a comparative approach