Blind Image Restoration via Fast Diffusion Inversion

Authors: Hamadi Chihaoui, Abdelhak Lemkhenter, Paolo Favaro

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We experimentally validate BIRD on several image restoration tasks and show that it achieves state of the art performance. Project page: https://hamadichihaoui.github.io/BIRD.
Researcher Affiliation Academia Hamadi Chihaoui Abdelhak Lemkhenter Paolo Favaro Computer Vision Group, Institute of Informatics, University of Bern, Switzerland {hamadi.chihaoui,abdelhak.lemkhenter,paolo.favaro}@unibe.ch
Pseudocode Yes Algorithm 1 BIRD: Image Restoration; Algorithm 2 DDIMReverse ( x T , δt)
Open Source Code Yes Project page: https://hamadichihaoui.github.io/BIRD. We include the training details and we provide the code in the Supplementary.
Open Datasets Yes We evaluate our method both on the validation datasets of Image Net [5] and Celeb A [10] at 256 256 pixel resolution.
Dataset Splits Yes We evaluate our method both on the validation datasets of Image Net [5] and Celeb A [10] at 256 256 pixel resolution. The PSNR mean and standard deviation are computed over 10 runs.
Hardware Specification Yes All experiments are carried out on a Ge Force GTX 1080 Ti.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions).
Experiment Setup Yes For all our experiments, we use δt = 100 and a maximum number of iterations N = 200. Adam is adopted as an optimizer with a 0.003 learning rate.