){kind=avatar}
Jeovane Honório Alves
Machine Learning Researcher
Biography
Jeovane Honório Alves holds a Ph.D. in Computer Science from the Federal University of Parana (UFPR). He is passionate about leveraging his research and professional interests in artificial intelligence (AI), computer vision (CV), machine learning (ML), deep learning (DL), neural architecture search (NAS), natural language processing (NLP), evolutionary computation, and image processing to drive innovation and improve business outcomes.
His doctoral research was focused on ML, CV, and NAS applied to natural and medical images, including CIFAR, ImageNet, and CHAOS challenge datasets. He has extensive experience working with several frameworks, including PyTorch, Numpy, Scikit-learn, Pandas, Keras, SciKit-Learn, ITK and VTK, Detectron, MMDetection, and OpenCV, to solve various tasks such as image classification, object detection, semantic and instance segmentation.
He is excited to bring his expertise and problem-solving skills to collaborate with talented teams and create innovative solutions that can transform industries and improve people’s lives.
Currently living in Brazil.
Download my resumé.
Download my PhD thesis.
Download my MSc dissertation.
- Data Science
- AutoML
- Computer Vision
-
PhD in Computer Science, 2021
Federal University of Parana (UFPR)
-
MSc in Electrical Engineering, 2016
Federal University of Parana (UFPR)
-
Technologist in Systems Analysis and Development, 2013
Federal University of Parana (UFPR)
Skills
Main Experience
- Research projects applied to Natural Language Processing and Computer Vision.
- Anomaly Detection (i.e., diseases and plagues) in agricultural applications;
- Development and Deployment of Deep Learning Models for computer vision tasks;
- Object Detection and Instance Segmentation using state-of-the-art models such as YOLOX, Cascade Mask-RCNN and Swin Transformer;
- Proficient in MMDetection framework for computer vision tasks;
- Utilization of Deep Active Learning methods for reducing annotation costs in computer vision tasks.
- Generation of custom Deep Learning (DL) models, including neural architecture search (NAS), for Computer Vision (CV) problems;
- Application of evolutionary computation techniques to optimize NAS;
- Image classification and segmentation of natural and volumetric (3D) medical images, such as CIFAR, ImageNet, and CHAOS challenge datasets;
- Expertise in PyTorch and its eager execution for dynamic NAS and model optimization;
- Extensive use of various convolutional operations, such as dilated, separable depthwise, and grouped convolutions, as well as auxiliary operations like batch normalization and pooling;
- Strong focus on data augmentation for image classification and segmentation;
- Utilization of optimization methods, such as SGD, Adam, cosine annealing, one-cycle scheduler, and automatic mixed-precision.
- Conventional feature engineering for image classification;
- Image processing techniques such as superpixels for region-of-interest segmentation;
- Medical image processing using the Insight Toolkit (ITK) in C++;
- Image processing using OpenCV;
- Machine learning using Scikit-Learn and Shark-ML;
- Classification methods such as Random Forest and SVM;
- Data storage using MongoDB.
