Slovakia

This work presents the computer vision and machine learn-

IWSSIP 2021

Simultaneous localization and mapping in 3Dspace

simultaneous localization and mapping

3d point cloud

lidar

We are pleased to announce the Thirty-Eighth AAAI Conference on Artificial Intelligence (AAAI-24), which will be held in Vancouver, British Columbia at the Vancouver Convention Centre – West Building from 20-27 February, 2024.

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. AAAI-24 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and a range of other activities to be announced.

We expect for AAAI-24 to be an in-person conference – one author of all accepted papers will be expected to present work in person unless there are exceptional circumstances that prevent this.<br><br><br><br>

In order to access the AAAI-24 event page you need to register [here](https://aaai.org/aaai-conference/registration/)

AAAI 2024

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. 

technical paper


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)

"Veterinary medicine emphasizes accidents caused by toxic plants with domestic animals as an extremely important topic, as the right diagnosis can be crucial for the affected animal. In this work, we propose the classification of toxic ornamental plants, according to nine different categories, using five widely-known CNN architectures, namely: DenseNet, ResNet, VGG16, VGG19 and Xception. The rationale behind it is that the automatic identification of these types of plant can be a useful tool to help in the prevention of those accidents. The authors have carefully curated a database to support the development of this work, collecting images available on the Pinterest website, and also performing some important data pre-processing. This database was also made available as a contribution of this work.
Transfer learning was employed by taking advantage of feature learned from the ImageNet dataset. We also analyzed the heat maps generated by the Layer-wise Relevant Propagation method, which allowed to observe the individual behavior of the best and worst architectures. The best performance was achieved using DenseNet, with an accuracy of 97.67%. That model managed to generalize very well, even to deal with noisy images, which are frequent in photos of decorative environments."

Classification of Toxic Ornamental Plants for Domestic Animals using CNN

The incidence of breast cancer has been increasing significantly worldwide. In 2020, more than 2.2 million new cases were registered. Breast cancer tends to alter the thermal pattern of the breasts. In this scenario, thermography appears as a functional examination capable of pointing out thermal changes. Several studies explore the use of this examination for cancer screening and diagnosis. However, few investigate its use for monitoring breast cancer treatment. This work aims at advancing these investigations using dynamic thermography. The data are clustered to obtain areas with higher temperatures and, from these, time series are built to compute different measures and follow up the tumor progress. The achieved results are promising and show that this approach could indicate the patient’s tumor evolution.

Clustering Thermography Data to Follow up Breast Cancer Treatment

"In recent years the native advertisement is becoming more and more prevalent in online spaces. Differentiating between genuine content and native advertisement using Natural Language Processing is therefore also becoming a very interesting research topic. In this paper, we examine the possibilities of using deep textual representation for the Slovak language to recognize the ""PR articles"" (that serve as a native advertisement in this context) from authentic news articles on popular Slovak news websites. We show that the BERT (Bidirectional Encoder Representations from Transformers) embeddings as a text representation are sufficient for this task (achieving accuracy over 80% even with a Logistic Regression model) and that the models generally perform better without prior lemmatization.
"

Automatic recognition of Native advertisements for the Slovak language

We create a compact 3D model-based gait feature out of

CNN based speech commands recognition

In this paper, a novel feature extraction approach for suddendeath prediction in patients with Chagas Disease (ChD) using electro-cardiogram  (ECG)  signals  is  proposed.  The  Structural  Co-OccurrenceMatrix (SCM) configuration is modified to extract the main frequenciesof ECG signals. Also, our approach uses only the QRS complex as a pa-rameter. This is the first time, to our knowledge, that SCM is been usedto solve this problem. F1-Score, precision, recall, accuracy, and compu-tational cost are employed as metrics for evaluating SCM performance.SCM with four well-known classifiers used in the literature are assessedand  the  results  are  compared  with  two  traditional  extractors  (Higher-Order Statistics (HOS) and Fourier Transform). Moreover, a dataset forfuture  studies  is  developed.  The  results  confirm  that  SCM,  combinedwith  frequency  domain  filtering,  can  extract  discriminative  structuralinformation on the task of interest.

Sudden Death Classification Using Electrocardiogram Signals and Structural Co-occurrence Matrix in Patients with Chagas Disease

In the presented contribution we propose motor imagery based BCI system for device control. Based on the EEG datasets we perform a two and three class classification of selected features for real motor movements and motor imagery. Features are created using complex wavelet transform. The classification is based on convolution neural networks. The results show that the method has a similar performance as the known reference method for given datasets.

An Approach for BCI using Motor Imagery Based on Wavelet Transform, and Convolutional Neural Network

Fingerprints are widely used for biometric validation worldwide. Since human fingerprints are unique and remain constant over time, it provides an easy-to-use, reliable, and economical authentication method. In addition to that, fingerprint recognition systems are of great importance because of their applicability in our lives. This work presents a classification methodology based on Henry Classification System using Convolutional Neural Networks (CNNs) models, such as Darknet, Alexnet, Resnet, VGG16, and Deep Belief Network. Besides that, we evaluate our proposal by carrying out experiments using grayscale images and pre-processed images as input on the classification step with the combination of the Gabor filter and the morphological thinning operation. We have obtained the highest result accuracy of 95.1% in the NIST Special Database 4, a widespread fingerprint dataset, using the Resnet 34 model in grayscale images. The proposed approach was evaluated with extraction strategies of classic attributes and based on convolutional networks. According to the results, the proposed methodology presents promising results, surpassing the traditional approaches present in the literature.

Fingerprint Classification based on the Henry System via ResNet

"The paper presents a multi-tool environment with an online assistant for Direct Digital Synthesis (DDS). The environment is integrated into the Online CADCOM platform and serves the Moodle-based Computer-Aided Design in Communication course in the Faculty of Telecommunications at the Technical University of Sofia. The design of DDS, as an all-digital signal generator, is one of the topics in the category “Oscillators design”. 
The environment integrates the online tools which can be used for DDS and the MATLAB (Simulink) -Vivado (FPGA) combination for creating an experimental setup.  The online assistant offers multimedia and interactive digital content on DDS, leading students in the implementation of the set of tools integrated for solving DDS design tasks. By combining teaching materials with tasks, quizzes, and forums the online assistant helps students develop critical thinking and encourages collaboration between them. Learning Management Systems, Moodle, and collaborative learning with an open web-based blog enhance students’ analytical skills as they share knowledge and discuss in a group. The new environment is probated with a group of Erasmus+ students on a mobility in TU-Sofia, following the CADCOM course.
"

Multi-tool environment with an online assistant for teaching Direct Digital Synthesis

"The main aim of this article is to compare U-Net models
with different methods of medical image segmentation in ophthalmology when searching for the optic disc and vessels on the retina. We
used freely available databases to analyze the results. DRIVE, STARE
and MESSIDOR were used for blood vessel segmentation and DRION
database was used for optic disc segmentation. In vascular segmentation,
we compared the method based on morphological operations with U-Net
convolutional network. The best results were obtained by neural network, where we achieved accuracy of 96.7% on DRIVE dataset. Methods
based on morphological operations, Attentional U-Net network and Fast
Radial Symmetry Transform were used for optic disc localization, where
the U-net network achieved the best sensitivity of 91.6%. We show that
in our tests U-Net architectures achieve better results in vessel and optic disc segmentation even compared to standard methods, that do not
depend on the amount of available data."

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

In the presentation we show the main aspects of the work. We made a brief introduction to the research and related works. We explain the methodology step by step and briefly discuss the results and conclusion.

Premium content

Downloads

Next from IWSSIP 2021

Classification of Toxic Ornamental Plants for Domestic Animals using CNN

Similar lecture

Multi-Constellation-Inspired Single-Shot Global LiDAR Localization