Moacir Antonelli Ponti

ORCID: 0000-0003-2059-9463
Publications
Citations
Views
---
Saved
---
About
Contact & Profiles
Research Areas
  • Advanced Image and Video Retrieval Techniques
  • Anomaly Detection Techniques and Applications
  • Domain Adaptation and Few-Shot Learning
  • Speech Recognition and Synthesis
  • Image Retrieval and Classification Techniques
  • Multimodal Machine Learning Applications
  • Neural Networks and Applications
  • Human Pose and Action Recognition
  • Machine Learning and Data Classification
  • Advanced Neural Network Applications
  • Speech and Audio Processing
  • Image Processing Techniques and Applications
  • Balance, Gait, and Falls Prevention
  • Music and Audio Processing
  • Electricity Theft Detection Techniques
  • Network Security and Intrusion Detection
  • Machine Learning and Algorithms
  • Advanced Vision and Imaging
  • Advanced Image Processing Techniques
  • Photoacoustic and Ultrasonic Imaging
  • Generative Adversarial Networks and Image Synthesis
  • Robotics and Sensor-Based Localization
  • Context-Aware Activity Recognition Systems
  • Evolutionary Algorithms and Applications
  • Image and Signal Denoising Methods

Brazilian Society of Computational and Applied Mathematics
2013-2024

Universidade de São Paulo
2015-2024

Universidade Federal de São Carlos
2005-2023

University of Surrey
2016

Deep Learning methods are currently the state-of-the-art in many Computer Vision and Image Processing problems, particular image classification. After years of intensive investigation, a few models matured became important tools, including Convolutional Neural Networks (CNNs), Siamese Triplet Networks, Auto-Encoders (AEs) Generative Adversarial (GANs). The field is fast-paced there lot terminologies to catch up for those who want adventure waters. This paper has objective introduce most...

10.1109/sibgrapi-t.2017.12 article EN 2017-10-01

Multiple classifier combination methods can be considered some of the most robust and accurate learning approaches. The fields multiple systems ensemble developed various procedures to train a set machines combine their outputs. Such have been successfully applied wide range real problems, are often, but not exclusively, used improve performance unstable or weak classifiers. In this tutorial presented basic terminology field, discussion on effectiveness algorithms, diversity concept, for...

10.1109/sibgrapi-t.2011.9 article EN 2011-08-01

Background: Cannabidiol (CBD) is one of the main components Cannabis sativa and has anxiolytic properties, but no study been conducted to evaluate effects CBD on anxiety signs symptoms in patients with Parkinson’s disease (PD). This aimed impacts acute administration at a dose 300 mg measures tremors induced by Simulated Public Speaking Test (SPST) individuals PD. Methods: A randomised, double-blinded, placebo-controlled, crossover clinical trial was conducted. total 24 PD were included...

10.1177/0269881119895536 article EN Journal of Psychopharmacology 2020-01-07

Sketchformer is a novel transformer-based representation for encoding free-hand sketches input in vector form, i.e. as sequence of strokes. effectively addresses multiple tasks: sketch classification, based image retrieval (SBIR), and the reconstruction interpolation sketches. We report several variants exploring continuous tokenized representations, contrast their performance. Our learned embedding, driven by dictionary learning tokenization scheme, yields state art performance...

10.1109/cvpr42600.2020.01416 article EN 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2020-06-01

In this paper, we propose SC-GlowTTS: an efficient zeroshot multi-speaker text-to-speech model that improves similarity for speakers unseen during training.We a speaker-conditional architecture explores flow-based decoder works in zero-shot scenario.As text encoders, explore dilated residual convolutional-based encoder, gated and transformer-based encoder.Additionally, have shown adjusting GAN-based vocoder the spectrograms predicted by TTS on training dataset can significantly improve...

10.21437/interspeech.2021-1774 article EN Interspeech 2022 2021-08-27

YourTTS brings the power of a multilingual approach to task zero-shot multi-speaker TTS. Our method builds upon VITS model and adds several novel modifications for training. We achieved state-of-the-art (SOTA) results in TTS comparable SOTA voice conversion on VCTK dataset. Additionally, our achieves promising target language with single-speaker dataset, opening possibilities systems low-resource languages. Finally, it is possible fine-tune less than 1 minute speech achieve similarity...

10.48550/arxiv.2112.02418 preprint EN other-oa arXiv (Cornell University) 2021-01-01

The development of low-cost remote sensing systems is important in small agriculture business, particularly developing countries, to allow feasible use images gather information. However, obtained through such with uncalibrated cameras have often illumination variations, shadows, and other elements that can hinder the analysis by image processing techniques. This letter investigates combination vegetation indices (color index extraction, visual index, excess green) mean-shift algorithm,...

10.1109/lgrs.2012.2193113 article EN IEEE Geoscience and Remote Sensing Letters 2012-07-11

Devices and sensors for identification of fallers can be used to implement actions prevent falls allow the elderly live an independent life while reducing long-term care costs. In this study we aimed investigate accuracy Timed Up Go test, fallers' identification, using fusion features extracted from accelerometer data. Single dual tasks TUG (manual cognitive) were performed by a final sample (94% power) 36 community dwelling healthy older persons (18 paired with 18 non-fallers) they wear...

10.1371/journal.pone.0175559 article EN cc-by PLoS ONE 2017-04-27

Image classification is one of the main research problems in computer vision and machine learning. Since most real-world image applications there no control over how images are captured, it necessary to consider possibility that these might be affected by noise (e.g. sensor a low-quality surveillance camera). In this paper we analyse impact three different types on descriptors extracted two widely used feature extraction methods (LBP HOG) denoising can help mitigate problem. We carry out...

10.48550/arxiv.1609.02781 preprint EN cc-by arXiv (Cornell University) 2016-01-01

Leukaemia is a dysfunction that affects the production of white blood cells in bone marrow. Young are abnormally produced, replacing normal cells. Consequently, person suffers problems transporting oxygen and fighting infections. This article proposes convolutional neural network (CNN) named LeukNet was inspired on blocks VGG-16, but with smaller dense layers. To define parameters, we evaluated different CNNs models fine-tuning methods using 18 image datasets, resolution, contrast, colour...

10.3390/s21092989 article EN cc-by Sensors 2021-04-24

In this paper we deal with the problem of feature selection by introducing a new approach based on Gravitational Search Algorithm (GSA). The proposed algorithm combines optimization behavior GSA together speed Optimum-Path Forest (OPF) classifier in order to provide fast and accurate framework for selection. Experiments datasets obtained from wide range applications, such as vowel recognition, image classification fraud detection power distribution systems are conducted asses robustness...

10.1109/icassp.2011.5946916 article EN 2011-05-01

Low cost remote sensing imagery has the potential to make precision farming feasible in developing countries. In this article, authors describe image acquisition from eucalyptus, bean, and sugarcane crops acquired by low-cost low-altitude systems. They use different approaches handle images both RGB NIR (near-infrared) bands estimate quantify plantation areas.

10.1109/mcg.2016.69 article EN IEEE Computer Graphics and Applications 2016-07-01
Coming Soon ...