This repository contains a collection of machine learning projects that are part of the Zero to Mastery TensorFlow course. Each project focuses on different problem domains and utilizes various techniques and models for analysis and prediction. The projects are implemented in Jupyter Notebook format (.ipynb) and organized in the notebooks folder. Additionally, there is a folder saved models which includes some pre-trained models I created. The repository also includes some Python (.py) files that contain useful functions for model training and evaluation, plotting learning curves and preprocessing of image-files.
-
Binary Classification using sklearn make_moons
This project explores binary classification using the
make_moonsdataset from the sklearn library. The objective is to build a machine learning model that accurately classifies data points into two categories. -
Fashion MNIST Classification
The objective of this project is to build a machine learning model capable of accurately classifying fashion images into their respective categories. The dataset used is the Fashion MNIST dataset, which consists of grayscale images of various clothing items. The project employs deep learning techniques, specifically convolutional neural networks, to achieve high accuracy in classification.
-
Transfer Learning with Feature Extraction and Fine-tuning
The project focuses on transfer learning using the 10% subset of the Food101 images dataset. The EfficientNet model is used, and experiments are conducted using feature extraction and fine-tuning techniques. Projects evaluate the model's performance and visualizes its most wrong predictions, providing insights into the model's strengths and weaknesses.
-
Text Binary Classification using Kaggle's Natural Language Processing with Disaster Tweets
This project compares different model architectures for text binary classification using the Kaggle's nlp_getting_started dataset. The objective is to determine whether a tweet is related to a disaster or not. The project implements several models, including Naive Bayes, feed-forward neural network, LSTM, GRU, bidirectional-LSTM, 1D Convolutional Neural Network, and TensorFlow Hub Pretrained Feature Extractor. The performance of each model is compared to identify the most effective approach.