TOC-SR: Task-Optimal Compact diffusion for Image Super Resolution
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866915978254745600 |
|---|---|
| author | Vajrala, Sowmya Bankar, Akshay Arveti, Manjunath Pandith, Shreyas Kodavanti, Sravanth Sanyal, Subhajit Unde, Amit Miriyala, Srinivas Soumitri |
| author_facet | Vajrala, Sowmya Bankar, Akshay Arveti, Manjunath Pandith, Shreyas Kodavanti, Sravanth Sanyal, Subhajit Unde, Amit Miriyala, Srinivas Soumitri |
| contents | Diffusion models have recently demonstrated strong performance for image restoration tasks, including super-resolution. However, their large model size and iterative sampling procedures make them computationally expensive for practical deployment. In this work, we present TOC-SR, a framework for building efficient one-step super-resolution models by first discovering a compact diffusion backbone. Starting from a sixteen-channel latent diffusion model, we construct parameter-efficient surrogate blocks using feature-wise generative distillation and perform architecture discovery using epsilon-constrained Bayesian Optimization to minimize model complexity while preserving generative fidelity. The resulting compact diffusion backbone achieves a 6.6x reduction in parameters and a 2.8x reduction in GMACs compared to the expanded diffusion model. We then adapt this backbone for super-resolution and distill the diffusion process into a single-step generator. Experiments demonstrate that the proposed approach enables efficient super-resolution while maintaining strong reconstruction quality. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_02767 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | TOC-SR: Task-Optimal Compact diffusion for Image Super Resolution Vajrala, Sowmya Bankar, Akshay Arveti, Manjunath Pandith, Shreyas Kodavanti, Sravanth Sanyal, Subhajit Unde, Amit Miriyala, Srinivas Soumitri Computer Vision and Pattern Recognition Artificial Intelligence Diffusion models have recently demonstrated strong performance for image restoration tasks, including super-resolution. However, their large model size and iterative sampling procedures make them computationally expensive for practical deployment. In this work, we present TOC-SR, a framework for building efficient one-step super-resolution models by first discovering a compact diffusion backbone. Starting from a sixteen-channel latent diffusion model, we construct parameter-efficient surrogate blocks using feature-wise generative distillation and perform architecture discovery using epsilon-constrained Bayesian Optimization to minimize model complexity while preserving generative fidelity. The resulting compact diffusion backbone achieves a 6.6x reduction in parameters and a 2.8x reduction in GMACs compared to the expanded diffusion model. We then adapt this backbone for super-resolution and distill the diffusion process into a single-step generator. Experiments demonstrate that the proposed approach enables efficient super-resolution while maintaining strong reconstruction quality. |
| title | TOC-SR: Task-Optimal Compact diffusion for Image Super Resolution |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2605.02767 |