Denoising: From Predicted Noise to an Estimated Image and One Reverse Step
This lesson is part of the course
In this lesson
You take the network's noise guess at t = 10, rebuild an estimate of the clean image, measure its error and take one reverse step back to t = 9.
Buy the course once for $9.99 to unlock every lesson and lab, the final exam and the certificate – lifetime access, no subscription.
Lessons in this course
- A Toy Image in Numbers: Pixel Values, Scaling to −1…1 and Back Locked
- Noise Schedules: Linear vs Cosine β and the Running Product ᾱ Locked
- Forward Noising: Mixing Image and Noise at t = 0, 5, 10 and 20 Locked
- Signal-to-Noise: SNR per Step, Decibels and Where the Image Disappears Locked
- Denoising: From Predicted Noise to an Estimated Image and One Reverse Step Locked
- Guidance: Mixing Prompt and No-Prompt Predictions with a Guidance Scale Locked
- Steps vs Time: Seconds per Image, Cost per 1,000 Images & Diminishing Returns Locked
- Latent Space: Pixel vs Latent Tensor Sizes, Compression Factor & Memory per Batch Locked