Udacity-deep-learning-projects

Projects from Udacity deep learning course

The Udacity Deep Learning course includes a series of hands-on projects designed to develop practical expertise in neural networks and deep learning techniques.

First project - MNIST Handwritten digits

The first project involves designing a custom neural network to tackle a classification task using the MNIST dataset of handwritten digits. The model is trained and evaluated to achieve optimal accuracy in recognizing digit patterns.

Second project - CNN - Algorithms for landmark classification

The second project, part of the convolutional neural networks (CNN) module, focuses on developing an algorithm for landmark classification.

1. The very first assignment is to design and train a CNN from scratch and get a test accuracy of at least 50%.

2. The second assignment is to use transfer learning and compare the results with the first assignment.

3. The third and last part will be about building a simple interactive app with the exported model.

Third project - Text Translation and Sentiment Analysis using Transformers

The third project aims to analyze sentiment in movie reviews across three different languages—English, French, and Spanish—using three distinct datasets. A pre-trained model is utilized to classify the sentiment and provide insights into language-specific sentiment trends.

Fourth project - Face generation with GAN

In the fourth project, a Generative Adversarial Network (GAN) is designed and trained on a dataset of facial images. The objective is to generate and store new, realistic face images that closely resemble human features.