Showcase
These projects represent a collection of my personal endeavors and projects, developed during my own time outside of professional work.
Eventum
Technologies
- Mobile: Typescript, React Native
- Backend: Typescript, NestJS, Supabase
Eventum is an app that lets you track various sources of events you’re interested in. With integrations spanning football, NBA, TV series, and more, it offers a comprehensive platform for staying updated.
Available at https://apps.apple.com/ar/app/eventum/id1455704418
QuotesBot
Technologies: Python
I developed a Wikiquote’s scraper, which I then used to create a bot that tweets famous quotes. This bot provides daily inspiration and wisdom from a range of historical and contemporary figures.
The bot is located here: https://twitter.com/QuoteBot17
ARZombies
Technologies: Unity, Augmented Reality
While following a udemy course, I made this ARZombies game: a first person survival shooter.
TGS Augmented Reality (AR) Expo
Technologies: React Native + Viro React (AR framework)
Augmented reality (AR) app that shows a portal to visualize the company surroundings. This portal consisted of 360 Images and videos of different locations, the user could then cycle through those locations. This project was showcased by TGS (www.tgs.com.ar) in Argentina Oil&Gas 2019 event. The app run in an iPad pro which was positioned at the stand of the company.
Object detection (Master Thesis)
Technologies: Python, Machine learning, Caffe
Title: Object Detection Based on Convolutional Neural Networks Trained on Synthetically Generated Data
The goal of my master thesis was to:
- Generate synthetic datasets representing different features of an object of interest by using a supplied 3d model. In this case, a volkswagen golf. This dataset was generated by using Blender and Blender’s bpy, which allows to things like taking pictures from different angles. The background images and car colors were also modified. The datasets generated were ~250k images (and in every case, automatically annotated).
- Train an object detection algorithm based on Convolutional Neural Network (CNN) with the generated dataset. The type of CNN used was SSD (SingleShot Multibox Detector). Once trained, these networks were tested with real world images and videos.
Here’s an example of object detection processed by the CNN trained with synthetic images:
- Combine the previously trained CNN with a pose estimation CNN, we then recreate the object in Blender:
You can check my thesis here
Brubank Dashboard
Technologies
- Frontend: Typescript, React
- Backend: Python, Flask
I made a dashboard while working at Brubank. This was a simple project and I did it alongside my daily work (which consisted in iOS development).
Chess (Hobby)
Technologies: Elm, Functional Programming
Chess is one of my favorite games, and I made a chess using Elm and functional programming. Tackling chess in a different paradigm was really interesting. It lead me to understand chess in a different way.
For example to calculate checkmate, obtain all the possible piece movements (posibble boards) that can be generated from that particular board and check if there’s no possible way of defending the king.
The code is live here https://github.com/FranDepascuali/FunctionalChess
You could check it here, but heroku took it down :( https://functionalchess.herokuapp.com/