The next is a humble collection of projects from my early days in Deep Learning. They were completed during a time when the domain of artificial neural networks was just exploding like a confetti cannon. I cannot help but feel a sense of accomplishment for overcoming the challenges of an era where GPUs were scarce and good digestible documentation was more elusive than Bigfoot. Each of these projects held a special meaning for me. For example, one allowed me to experience the process of creating a product from the idea conception to the prototype presentation in front of more than 200 people. Another led my current boss to discover and recruit me for his startups REstyle and DPlace. And another inspired me to start The Almendra Project, and so on. Now, I have significantly more experience and knowledge to create new ones, which is so excing. Working on getting a new batch of projects out soon...
Imagine being able to converse with one of the oldest indigenous communities in Mexico, the Wixaricás, better known as Huicholes, whose remote location in the Sierra Madre Occidental has made access to new technologies nearly impossible, so new generations are being forced to leave their homeland and move to the big cities, putting their culture at risk of losing their mother tongue. Mariya is the solution to bridging the communication gap. As a translation device, Mariya allows you to engage with the Huicholes in their native language, opening doors to a world of cultural exchange and understanding. It aims to preserve this precious piece of cultural history and provides you with a once-in-a-lifetime opportunity to connect with their culture in their own language.
Discover more projects on my GitHub profile.
Have you ever wondered how a new hairstyle would transform your look? Imagine being able to try on any hairstyle without the risk of it turning out wrong, no more guessing or relying on magazine images that could lead to false expectations. This project identifies your hair's location in photos or videos; Which is the first step in creating a virtual hair "try-on" mirror leaving you feeling confident and empowered of trying new hair styles.
Did you know that skin cancer is the most common cancer in the world? In the U.S. alone, there are 5.4 million cases every year. Melanoma is the most dangerous type, and its early detection is key to surviving it since it can increase your survival chances by 99-100%. This project is designed to help diagnose melanoma by distinguishing it from benign skin lesions like nevus and seborrheic keratosis.
Pixelwise segmentation allows autonomous cars to figure out where they can go and where they can't go, helping them see the world with laser-sharp precision and avoid obstacles by navigating through the available space. This project used the CamVid dataset to create a neural network able to classify and accurately locate, pixel by pixel, 32 different objects.
If you've ever tried to find your keys in a messy room, you know the struggle of locating an object. Well, imagine a computer that can do it in a snap! You only look once (YOLO) is like a superhero with super speed for computer vision. It can recognize and locate objects in real-time, making it perfect for applications that require quick decision-making, for example, in self-driving cars to quickly detect pedestrians suddenly crossing the street. This is an implementation of YOLO v3.