Skip to content

Deep Learning Summer School + Tensorflow + OpenCV cascade training + YOLO + COCO + CycleGAN + AWS EC2 Setup + AWS IoT Project + AWS SageMaker + AWS API Gateway + Raspberry Pi3 Ubuntu Core

Notifications You must be signed in to change notification settings

Micseb/Repo-2018

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Python Codes: IoT, Machine Learning, Computer Vision

Codes in Machine Learning, Deep Learning, Reinforcement Learning, Artificial Intelligence and Computer Vision

Welcome to my GitHub repo.

I am a Data Scientist and I code in Python. Here you will find some Machine Learning, Deep Learning, Natural Language Processing, Artificial Intelligence and Computer Vision models I developed.

Data for models available at:

https://drive.google.com/drive/folders/0B0RLknmL54khU2UwX3dnX1E1WHc?usp=sharing

Keras version used in models: keras==2.0.8 Tensorflow version used in models: tensorflow==1.7.0

AWS API Gateway

This folder presents general guidelines to create an endpoint and API key in AWS API Gateway, that will receive data from a Python notebook sending sensor data, that arrives in API Gateway, is processed and cleaned in Lambda and is stored in DynamoDB. Data is then available for S3 to create a data pipeline and further visualization with Kibana or Quick Sight.

AWS SageMaker

This sector presents the development of a Docker container with a customized Machine Learning model that is pushed into Elastic Container Registry and generates and endpoint in AWS SageMaker.

COCO Model

COCO (Common Objects in Context) is an image segmentation model. This folder presents my Pull Request regarding pycocotools library incompatibility between Python 2 and 3 suggesting a fix.

CPU Temperature - IoT Project at AWS

In this project I turned my notebook into an IoT device, where CPU temperature is collected via Linux command line run inside a Python notebook and then sent via MQTT protocol to Amazon AWS IoT service, integrated with DynamoDB, Data Pipeline, S3, Quick Sight and Cloud Watch. An alternative solution for near real-time data is also presented, using Kinesis, Firehose, Elasticsearch and Kibana.

CycleGAN Project - GANs

In this folder I presente the steps to train a CycleGAN for Style Transfer applied to paintings of Vincent van Gogh and Pablo Picasso using an EC2 GPU Tesla K-80. Examples of outputs are presented.

DEEP LEARNING SUMMER SCHOOL - GANs

In this folder you will find Tensorflow and Keras codes and also a Powerpoint presentation about GANs I developed for my lecture and workshop at the first Deep Learning Summer School in Brazil, at Goiás.

EC2 INSTANCE SETUP

A set up Manual I developed to create EC2 instances in Amazon AWS

FLASK

This folder contains the Python code for the REST API and also the request code.

KERAS

GANs: Generative Adversarial Networks

DCGANs: Deep Convolutional Generative Adversarial Networks

NLP

Notebooks presenting Word2Vec similarities in trained Wikipedia corpus, Portuguese language.

OPENCV

In this folder you will find a guide to create your own haarcascade.xml so that you can identify any object using OpenCV.

Raspberry Pi3 - IoT Project

In this project I use Raspberry Pi3 running in Ubuntu Core OS and attached a sensor and then send telemetry data to Amazon AWS IoT service, integrated with DynamoDB, S3, Quick Sight and Cloud Watch. I also plan to deploy a Deep Learning trained model into the device for classification purposes.

TENSORFLOW

Model for Regression

Model for Classification

LSTM: Long Short Term Memory Neural Networks

GANs: Generative Adversarial Networks

DCGANs: Deep Convolutional Generative Adversarial Networks

InfoGANs Mutual Information Adversarial Networks where loss function is customized. Mutual Info = H(B)-Sum(P(B=b).H(A|B=b)

Tensorboard: visualization of Tensorflow models' training

SEG-NET : Image Segmentation with SEG-NET

SMOTE : Synthetic Minority Oversampling Technique for imbalanced classes

YOLO Model

The YOLO Model (You Only Look Once) is a Deep Learning project for Real-Time object detection. Examples of outputs are presented.

Datasets available at Repo-2017

About

Deep Learning Summer School + Tensorflow + OpenCV cascade training + YOLO + COCO + CycleGAN + AWS EC2 Setup + AWS IoT Project + AWS SageMaker + AWS API Gateway + Raspberry Pi3 Ubuntu Core

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 70.8%
  • Python 28.8%
  • Dockerfile 0.4%