Lesson 3 • 2 min

The Decoder

Back to pixels

Unzipping a file

When you unzip a file, you get back all the original data from the compressed version. The VAE decoder takes the latent representation and reconstructs a full-resolution image.

Watch the decoder expand latent space to full resolution

The decoder is a learned neural network—not a simple algorithm. It was trained on millions of images to learn how to reconstruct fine details from the compressed representation.

VAE encode/decode

# The full VAE pipeline
class VAE:
    def encode(self, image):
        # 1024×1024×3 → 128×128×4
        return self.encoder(image)

    def decode(self, latent):
        # 128×128×4 → 1024×1024×3
        return self.decoder(latent)

# In generation:
# 1. Start with noise in latent space
# 2. Denoise in latent space (all the diffusion magic)
# 3. Decode final latent to pixels
pixels = vae.decode(denoised_latent)

Quick Win

Module 5 complete! You understand VAEs: compression to latent space for efficient processing, then decoding back to pixels.

Next Lesson