Improved Image Generation via Sparse Modeling

Roy, Ganz,; Elad, Michael

doi:10.48550/arxiv.2104.00464

Cited by 1 publication

(1 citation statement)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, Mahdizadehaghdam et al (2019) exploits patch-based sparsity and takes in a pre-trained dictionary to assemble generated patches. Ganz & Elad (2021) explores convolutional sparse coding in generative adversarial networks, arguing that the generator is a manifestation of the convolutional sparse coding and its multi-layered version synthesis process. Both methods have shown that using sparsity-inspired networks improves the image quality of GANs.…”

Section: Connections To Related Workmentioning

confidence: 99%

Closed-Loop Transcription via Convolutional Sparse Coding

Dai¹,

Chen²,

Tong³

et al. 2023

Preprint

View full text Add to dashboard Cite

Autoencoding has achieved great empirical success as a framework for learning generative models for natural images. Autoencoders often use generic deep networks as the encoder or decoder, which are difficult to interpret, and the learned representations lack clear structure. In this work, we make the explicit assumption that the image distribution is generated from a multi-stage sparse deconvolution. The corresponding inverse map, which we use as an encoder, is a multi-stage convolution sparse coding (CSC), with each stage obtained from unrolling an optimization algorithm for solving the corresponding (convexified) sparse coding program. To avoid computational difficulties in minimizing distributional distance between the real and generated images, we utilize the recent closed-loop transcription (CTRL) framework that optimizes the rate reduction of the learned sparse representations. Conceptually, our method has high-level connections to score-matching methods such as diffusion models. Empirically, our framework demonstrates competitive performance on large-scale datasets, such as ImageNet-1K, compared to existing autoencoding and generative methods under fair conditions. Even with simpler networks and fewer computational resources, our method demonstrates high visual quality in regenerated images. More surprisingly, the learned autoencoder performs well on unseen datasets. Our method enjoys several side benefits, including more structured and interpretable representations, more stable convergence, and scalability to large datasets. Our method is arguably the first to demonstrate that a concatenation of multiple convolution sparse coding/decoding layers leads to an interpretable and effective autoencoder for modeling the distribution of large-scale natural image datasets. 1

show abstract

Section: Connections To Related Workmentioning

confidence: 99%