Applications of Autoencoders — Part 4
Image Colorization with Deep Convolutional Autoencoders
Convert grayscale images to RGB color images
Today, we will work on the fourth practical application of autoencoders. Previously, we worked on the following types of applications.
- Image generation with autoencoders
- Image compression/dimensionality reduction with autoencoders
- Image denoising with autoencoders
Today, we will perform image colorization with autoencoders. In other words, we add color to grayscale images. After adding colors, the grayscale images will become RGB color images.
What are autoencoders?
Autoencoders (AEs) are a type of neural network architecture that consists of three key elements: Encoder, Latent Vector and Decoder. The encoder and decoder are non-linear functions. A given input is transformed into the latent vector by the encoder. The latent vector is the output of the encoder and will be the input for the decoder. The decoder recovers the input from the latent vector. The output of the decoder is approximately close to the original input.
