Geovane P.

Geovane P.

Solutions Architect at BriteCore | Senior Software Engineer | Full Stack Developer | M.Sc.

Rio de Janeiro , Brazil

Experience: 12 Years

Geovane

Rio de Janeiro , Brazil

Solutions Architect at BriteCore | Senior Software Engineer | Full Stack Developer | M.Sc.

130000 USD / Year

  • Notice Period: 15 Days

12 Years

Now you can Instantly Chat with Geovane!

About Me

Master's Degree in Computational Modeling and also Computer Engineer by Rio de Janeiro State University (UERJ). In addition to that, I am also a Technician in Industrial Informatics by the Federal Center of Technological Education Celso Suckow...

Show More

Skills

Development Tools

Networking & Security

Portfolio Projects

Automatic Face Recognition Using Digital Image Processing

https://github.com/geovanecomp/Automatic-face-recognition-using-digital-image-processing-

Company

Automatic Face Recognition Using Digital Image Processing

Role

Software Architect

Description

Automatic Face Recognition Using Digital Image Processing.

The objective of this project is to apply Image Processing algorithms to improve the image quality, distributing uniformly the image intensity, and removing noises. Then, analyze how Facial Recognition Methods would recognize faces in different noise conditions.

The image sources are very important and are directly related to the success rate of these recognition methods. Images that are naturally perfect, which means no noises and good luminosity, tend to have better results. However, noisy images need to be fixed or prepared before being applied in the recognition methods to achieve a better success rate.

The following example shows how important image processing is when applied into noisy images:

 

As you can see, the image on the right after being processed by the Histogram Equalization is clearer than before.

Because of that, multiple image processing methods were implemented, and due to that, would be very difficult to create distinct combinations between them. So, to make it easier, was implemented the Design Pattern Chain of Responsibility, making the addition of new methods or the combination between them extremely faster and simple.

Through this Design Pattern, four image databases were created based on the combination of the Image Processing Algorithms: Histogram Equalization, Laplacian Filter, Laplacian Filter with Suavization Filter, and finally, all these filters in the same order that were presented.

Each processed database was submitted to two face recognitions methods, the Normalized Correlation and Eigenfaces.

The new databases were created using the FEI database (http://fei.edu.br/~cet/facedatabase.html) as original source. FEI database is composed of images with high quality and controlled environment, considering several faces in different positions and luminosity conditions. 

As result of the facial methods and each combination of image processing, the average success rate are:

 

In this case, the bests images processing methods utilized were the Laplacian with Suavization Filter and Histogram Equalization.

This is a very short resume of my final Computer Engineer Degree project from Rio de Janeiro State University (UERJ). Feel free to send me an email (geovane.pacheco99@gmail.com) for questions, critiques, or compliments.

Show More Show Less

(2 Weeks Hiring Project): BriteCore Insurance System

https://github.com/geovanecomp/BriteCore-Insurance-System

Company

(2 Weeks Hiring Project): BriteCore Insurance System

Role

Full-Stack Developer

Description

BriteCore-Insurance-System

Welcome to BriteCore Insurance System. The project's goal is to build in one week a system that enables Insurers to define their own data model, and for that, was provided a fully tested API to allow interaction by any frontend system, such as VueJS.

In summary, this project was created using MySQL to create a dynamic data model, Django Restful framework for the backend API, and the frontend was built using the modern javascript framework VueJS.

For that, was created two internal projects, one called by backend and another called by frontend, and the communication between them was done through the REST pattern. The DevOps base was built using Docker, Docker-compose, and Amazon AWS to deploy it.

What was done

Data

Two data models were created, one that was implemented as an MVP ( Minimum Viable Product) and another to be the best approach, but consequently, would spend more time.

Here are (The MVP and the best approach): *(My personal bonus)

  • https://github.com/geovanecomp/BriteCore-Insurance-System/blob/master/docs/model/Implemented MVP.jpeg
  • https://github.com/geovanecomp/BriteCore-Insurance-System/blob/master/docs/model/Best approach model.jpeg

Backend

  • The backend was built using the community best practices and TDD oriented, with that, the API was fully tested (around 50 tests!). The API was created using the framework Django 2.0 and python 3.6.

  • The API is prepared for all HTTP verbs / REST. It means anyone is able to request a full CRUD of risk types, risk, field types, fields, and fields by risk. *(My personal bonus)

Here are a few localhost API screens:

  • https://github.com/geovanecomp/BriteCore-Insurance-System/blob/master/docs/api/api_root_page.jpeg
  • https://github.com/geovanecomp/BriteCore-Insurance-System/blob/master/docs/api/fields_by_risk_list.jpeg
  • https://github.com/geovanecomp/BriteCore-Insurance-System/blob/master/docs/api/structure_documentation.jpeg
  • https://github.com/geovanecomp/BriteCore-Insurance-System/blob/master/docs/api/api_interaction.jpeg

Frontend

  • The frontend was built using the modern javascript framework Vuejs, HTML5, CSS3 / Sass.

  • Three pages were created, one to make all CRUD API requests for Risk Types, one to simulate how to create the riks and their fields (almost finished), and finally the main page, which gets all risk fields and shows them in an appropriate way. *(My personal bonus)

Here are a few frontend screens:

  • https://github.com/geovanecomp/BriteCore-Insurance-System/blob/master/docs/system/welcome_page.jpeg
  • https://github.com/geovanecomp/BriteCore-Insurance-System/blob/master/docs/system/risk_type_crud.jpeg
  • https://github.com/geovanecomp/BriteCore-Insurance-System/blob/master/docs/system/manage_risks.jpeg
  • https://github.com/geovanecomp/BriteCore-Insurance-System/blob/master/docs/system/main_screen.jpeg

Devops *(My personal mega bonus)

  • The project was built using Docker and Docker-compose, with that anyone can make an easy and faster installation using these containers. With that, for example, was really easy to deploy the application to the AWS (Amazon Web Services).

How to execute a developer version

  1. Download and install Docker (Version 3);
  2. Download and install Docker-compose;
  3. Clone this project;
  4. From the project root folder execute to build and up the containers: docker-compose -f docker-compose.yml -f docker-compose.development.yml up --build;
  5. Access the backend container and execute the script db_initialize.sh to populate the database: docker-compose -f docker-compose.yml -f docker-compose.development.yml exec backend bash -c "sh db_initialize.sh";
  6. Access http://localhost:8080 to open the software or http://localhost:8000 to open the API documentation.
  7. Finally, have fun =).

How to execute the production version (Temporary Off)

  1. Just access: http://ec2-54-198-232-208.compute-1.amazonaws.com/

For questions, critiques or compliments, send me an email: geovane.pacheco99@gmail.com

Show More Show Less

Company

General Studying Repo

Role

Full-Stack Developer

Description

This repository is where I store all kind of courses and technologies that I am learning.
It is also where I can find practical examples of several areas.

Show More Show Less

Master's Degree: Energy Estimation for Particle Collisions

https://github.com/geovanecomp/Masters-Degree

Company

Master's Degree: Energy Estimation for Particle Collisions

Role

Software Architect

Description

The description and reproduction of physical phenomena are a common challenge in the field of computational modeling. In the context of high-energy calorimetry, the particle energy produced from collisions is absorbed and sampled in the form of a digitized signal.

In a high event-rate and luminosity conditions, the signal pile-up effect may arise due to the high occupancy of the detector’s readout channels, causing the distortion of the expected signal. In this context, this work evaluates the performance of the method known as Matched Filter, by applying it in two distinct approaches called Deterministic Matched Filter and Stochastic Matched Filter.

These approaches will be compared to the current method applied in the ATLAS Tile Calorimeter (TileCal), known as Optimal Filter. For the efficiency analysis, a computational environment was created to contain simulated data considering different signal pile-up conditions in the TileCal. Furthermore, a hybrid environment was also created, consisting of signals whose amplitudes are previously known and added to different conditions of pile-up noise. For both analyses, different signal-noise ratio conditions were considered.

The results show that the Stochastic Matched Filter presented high accuracy in estimating the amplitude, becoming an alternative for the TileCal energy estimation task.

Show More Show Less