Skip to content
View hugozanini's full-sized avatar
🎯
Focusing
🎯
Focusing

Block or report hugozanini

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Please don't include any personal information such as legal names or email addresses. Maximum 100 characters, markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
hugozanini/README.md

Hi, I'm Hugo Zanini!

Electrical Engineer @ UFMG

Tech Product Manager @ Nu







Linkedin: hugozanini
Linkedin: hugozanini

My projects

Image Animation with Generative AI

Human image animation has gained a lot of attention given its potential application across various domains, including social media, the movie industry, and entertainment.

In this work, I adapted the Magic Animate model using Detectron and stablediffusion-1.5 for running the pipeline in a colab notebook using only free computing resources!

πŸ’» Code

Qries

Custom YOLOv7 Object Detection with TensorFlow.js

This article shows how to train a custom YOLOv7 model run it directly in the browser in real-time and offline with TensorFlow.js.

This is the world's first open-source implementation of the YoloV7 model, a state-of-the-art object detector, in TensorFlow.js.

πŸ“° Article
πŸ’» Code
πŸ‘¨β€πŸ’» Live Demo

Qries

Real-time clothes segmentation with TF.JS

While semantic segmentation is already widely used in self-driving cars, there's untapped potential for it in the hashtag Fashion Industry too!

In this work I developed a segmentation model using Tensorflow.js, Transformers and lightweight multilayer perception (MLP) decoders for real-time real-time clothes segmentation in the browser:

πŸ’» Code

Qries

Realtime and offline SKU detection in the browser using TensorFlow.js

Recognizing SKUs in densely packed scenes is a well-known and challenging problem in the Consumer Packaged Goods industry. In this post on the Tensorflow blog, I showed how to train and deploy a model to solve it in real-time and offline in the browser of any device.

πŸ‘¨β€πŸ’» Live Demo
πŸ’» Code

Qries

Custom object detection in the browser using TensorFlow.js

Read my post on TensorFlow blog on how to train a MobileNetV2 using the TensorFlow 2 Object Detection API and Google Colab, convert the model, and run real-time inferences in the browser through TensorFlow.js.

πŸ‘¨β€πŸ’» Code

In the post, you will find all the steps to create a high-performance solution - from the data preparation until the deployment. There is no need for powerful computers or expensive graphic cards. It's possible to recognize up to 90 classes, and produce results as shown below:


TensorFlow Show and Tell

Watch my last talk about Semantic Segmentation in the Official TensorFlow YouTube Channel :)

NLP for sentiment Analysis in Portuguese

How to retrain BERT to perform real-time sentiment analysis in Portuguese using Pytorch, Python and javascript.

πŸ“° Model Training
πŸ“° Productizing the model
πŸ‘¨β€πŸ’» Code


Realtime semantic segmentation in the browser

Perform semantic segmentation directly from your browser using TensorFlow.js.
There is no need for complex libraries or powerful computers.

πŸ“° Article
πŸ‘¨β€πŸ’» Code

This project was selected as TensorFlow Community Spotlight Winner


Few-shot learning: Creating a real-time object detector using TensorFlow, Python and JavaScript

Take a few pictures of an object using your webcam, label the images and train a few-shot learning model using this Google Colab Notebook.

If you want to know more, go to the project repository.


In browser workout performance measurement using Pose Estimation

Use Tensorflow.js and PoseNet to measure your workouts directly from the browser.

πŸ“° Article: https://lnkd.in/eZmZ4GX
πŸ‘¨β€πŸ’» Code: https://lnkd.in/eUvbVB8


Creating a real-time object detector using tensorflow.js

This is a tutorial inspired by a Google I/O session, where I show how to use the pre-trained COCO-SSD model to create a real-time object detector in the browser using tensorflow.js.

πŸ‘¨β€πŸ’» Codelab: https://hugozanini.github.io/tf-object-detection


Markitt

Startup that solved data tracking and inventory control problems in coffee warehouses using Radio Frequency and Computer Vision.


Instagram Bot

Selenium-based bot to scrape data from Instagram posts.

Access the repo here


Fitting a sine function using a Neural Network

In this repository, you will find the codes to create a neural network capable of approximate a sine function.

All the explanations and discussions about creating the model and overfittings problems it can have are in the notebook.

In the end, it is expected that you produce results as shown below:

Pinned Loading

  1. yolov7-tfjs yolov7-tfjs Public

    Object Detection using Yolov7 in tensorflow.js

    JavaScript 82 20

  2. TFJS-object-detection TFJS-object-detection Public

    Real-time custom object detection in the browser using tensorflow.js

    JavaScript 128 68

  3. realtime-semantic-segmentation realtime-semantic-segmentation Public

    Implementation of RefineNet to perform real time instance segmentation in the browser using TensorFlow.js

    JavaScript 122 26

  4. Activities-Pose-Estimation Activities-Pose-Estimation Public

    In browser workout performance measurement using Pose Estimation and TensorFlow.js

    JavaScript 36 8

  5. sentiment-analysis sentiment-analysis Public

    Jupyter Notebook 18 9

  6. realtime-sku-detection realtime-sku-detection Public

    Python 16 6