PnP Inversion: Boosting Diffusion-based Editing with 3 Lines of Code

Authors: Xuan Ju, Ailing Zeng, Yuxuan Bian, Shaoteng Liu, Qiang Xu

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

Reproducibility Variable Result LLM Response
Research Type Experimental To assess image editing performance, we present PIE-Bench, an editing benchmark with 700 images showcasing diverse scenes and editing types, accompanied by versatile annotations and comprehensive evaluation metrics. Compared to state-of-the-art optimization-based inversion techniques, our solution not only yields superior performance across 8 editing methods but also achieves nearly an order of speed-up.
Researcher Affiliation Collaboration Xuan Ju1,2 , Ailing Zeng2 , Yuxuan Bian1, Shaoteng Liu1, Qiang Xu1 1The Chinese University of Hong Kong (CUHK) 2International Digital Economy Academy (IDEA)
Pseudocode Yes Algorithm 1: Real Image Editing Pipeline with Pn P Inversion
Open Source Code Yes 2Code is available at https://github.com/cure-lab/Pn PInversion.
Open Datasets No No, the paper introduces PIE-Bench, a benchmark dataset of 700 images, and describes its construction in detail. However, it does not provide a direct URL, DOI, or specific repository name for accessing the dataset itself.
Dataset Splits No No, the paper introduces the PIE-Bench dataset and evaluates performance on it, but it does not provide specific details on how the dataset is split into training, validation, and test sets for model training or evaluation.
Hardware Specification Yes We perform the inference of different editing and inversion methods in the same setting unless specifically clarified, i.e., on RTX3090 following their open-source code with a base model of Stabe Diffusion v1.4 in 50 steps, with an inverse guidance scale of 1 and a forward guidance scale of 0. ... We test inference time per image of different inversion techniques and Prompt-to Prompt (Hertz et al., 2023) on one NVIDIA A800 80G to evaluate efficiency.
Software Dependencies No No, the paper mentions using 'Stable Diffusion v1.4' as a base model but does not provide specific version numbers for software dependencies like Python, PyTorch, or CUDA.
Experiment Setup Yes We perform the inference of different editing and inversion methods in the same setting unless specifically clarified, i.e., on RTX3090 following their open-source code with a base model of Stabe Diffusion v1.4 in 50 steps, with an inverse guidance scale of 1 and a forward guidance scale of 0.