Saved in:
| Main Author: | |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2506.07286 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916784881270784 |
|---|---|
| author | Chakravarty, Aditya |
| author_facet | Chakravarty, Aditya |
| contents | Diffusion models have shown remarkable flexibility for solving inverse problems without task-specific retraining. However, existing approaches such as Manifold Preserving Guided Diffusion (MPGD) apply only a single gradient update per denoising step, limiting restoration fidelity and robustness, especially in embedded or out-of-distribution settings. In this work, we introduce a multistep optimization strategy within each denoising timestep, significantly enhancing image quality, perceptual accuracy, and generalization. Our experiments on super-resolution and Gaussian deblurring demonstrate that increasing the number of gradient updates per step improves LPIPS and PSNR with minimal latency overhead. Notably, we validate this approach on a Jetson Orin Nano using degraded ImageNet and a UAV dataset, showing that MPGD, originally trained on face datasets, generalizes effectively to natural and aerial scenes. Our findings highlight MPGD's potential as a lightweight, plug-and-play restoration module for real-time visual perception in embodied AI agents such as drones and mobile robots. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_07286 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Multi-Step Guided Diffusion for Image Restoration on Edge Devices: Toward Lightweight Perception in Embodied AI Chakravarty, Aditya Computer Vision and Pattern Recognition Machine Learning Robotics Diffusion models have shown remarkable flexibility for solving inverse problems without task-specific retraining. However, existing approaches such as Manifold Preserving Guided Diffusion (MPGD) apply only a single gradient update per denoising step, limiting restoration fidelity and robustness, especially in embedded or out-of-distribution settings. In this work, we introduce a multistep optimization strategy within each denoising timestep, significantly enhancing image quality, perceptual accuracy, and generalization. Our experiments on super-resolution and Gaussian deblurring demonstrate that increasing the number of gradient updates per step improves LPIPS and PSNR with minimal latency overhead. Notably, we validate this approach on a Jetson Orin Nano using degraded ImageNet and a UAV dataset, showing that MPGD, originally trained on face datasets, generalizes effectively to natural and aerial scenes. Our findings highlight MPGD's potential as a lightweight, plug-and-play restoration module for real-time visual perception in embodied AI agents such as drones and mobile robots. |
| title | Multi-Step Guided Diffusion for Image Restoration on Edge Devices: Toward Lightweight Perception in Embodied AI |
| topic | Computer Vision and Pattern Recognition Machine Learning Robotics |
| url | https://arxiv.org/abs/2506.07286 |