Completed
Comparison to denoising with deep image prior [Ulyanov et al., '18]
Class Central Classrooms beta
YouTube videos curated by Class Central.
Classroom Contents
Deep Decoder: Concise Image Representations from Untrained Networks - Lecture 2
Automatically move to the next video in the Classroom when playback concludes
- 1 Deep Decoder: Concise Image Representations from Untrained Networks Lecture 2
- 2 Recovering images from few data requires a model for natural images
- 3 Models for Natural Images: Wavelets + Sparsity
- 4 Models for Natural Images: Sparse Coding
- 5 Models for natural images: neural nets trained on large datasets
- 6 This talk: Untrained neural nets as a model of natural images
- 7 The deep decoder
- 8 Compression
- 9 Image compression
- 10 In contrast to deep decoder, other neural net architectures are complicated
- 11 Solving inverse problems with the deep decoder
- 12 Inverse problem
- 13 Image recovery with models
- 14 Denoising performance
- 15 Deep decoder is on par with state of the art for denoising
- 16 Why does the deep decoder work?
- 17 Why does the deep decoder denoise so well?
- 18 The deep decoder
- 19 Theory: Deep Decoder can only fit so much noise
- 20 Denoising rates
- 21 Proof
- 22 Deep image prior [ Ulyanov et al., '18]
- 23 Comparison to denoising with deep image prior [Ulyanov et al., '18]
- 24 How can linear upsampling, ReLUs, and liner combinations synthesize images efficiently?
- 25 Summary
- 26 Q&A
