Generating Flower Images Using Deep Convolutional GANs
Implementation of CNN-based GANs (DCGANs) with Keras to generate natural-looking, realistic flower images
13 min readApr 13, 2023
GANs (Generative Adversarial Networks) can generate natural-looking and more realistic things such as images, videos, sounds, and texts from the data in the given domain.
Deep convolutional GANs (DCGANs) perform really well when working with image data because convolutional layers are able to capture spatial relationships in image data.
In DCGANs, we use convolutional (downsampling) layers in the discriminator part and transposed convolutional (upsampling) layers in the generator part.
Today, in this article, we will discuss how to build and train a DCGAN model to generate natural-looking, realistic flower images using the flowers dataset that comes with TFDS (TensorFlow Datasets).
Prerequisites (must-read)
-------------------------
1. A Short Introduction to GANs in Generative Deep Learning
2. Using Inbuilt Datasets with TensorFlow Datasets (TFDS)
3. Convolutional Neural Network (CNN) Architecture
4. Coding a Convolutional Neural Network (CNN) Using Sequential API
5. Batch Normalization Explained in Plain English
6. Batch Size, Epochs and Training Steps in a Neural…