video
![Lesson thumbnail for Autoencoders for Image-to-Latent Compression for Diffusion Models](/assets/jpg/e98a1a1a.jpg)
Deep Learning Course - Level: Advanced
Now we know how the text portion of input is processed before being passed to U-Net during training and inference. Let's now explore how the image portion of input is processed.
Before passing the input image to U-Net, we first pass it to a pre-trained variational autoencoder to compress the image into a latent.
Committed by on