The "Generative Deep Learning with TensorFlow" course covers various topics and techniques in generative deep learning using TensorFlow.
Here's what you'll learn:
- Neural Style Transfer: You'll learn how to use transfer learning to combine the content of one image with the style of a painting, creating a new image that merges both.
- AutoEncoders: Starting with simple AutoEncoders on the MNIST dataset, you'll progress to more complex deep and convolutional architectures on the Fashion MNIST dataset. You'll explore the differences between DNN and CNN AutoEncoder models, learn how to denoise images and build a CNN AutoEncoder using TensorFlow to generate clean images from noisy ones.
- Variational AutoEncoders (VAEs): You'll cover VAEs, which allow for generating entirely new data. You'll specifically learn how to generate anime faces and compare them to reference images using VAEs.
- Generative Adversarial Networks (GANs): This section covers the invention, properties, and architecture of GANs, highlighting the differences from VAEs. You'll understand the roles of the generator and discriminator within the GAN model, the concept of two training phases, and the importance of introduced noise. Lastly, you'll build your own GAN capable of generating faces.
Overall, this course focuses on giving learners more control over model architecture and the tools to create and train advanced machine learning models using TensorFlow.
🆓 Free to Audit
🕒 Approx. 16 Hours
🧠 Advanced Level
🧾 Paid Certificate Available Upon Completion
🎓 Offered by DeepLearning.AI via Coursera
