Conditional Score Guidance for Text-Driven Image-to-Image Translation

Authors: Hyunsoo Lee, Minsoo Kang, Bohyung Han

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Through comprehensive experiments, we demonstrate that our approach achieves outstanding image-to-image translation performance on various tasks.
Researcher Affiliation Academia Hyunsoo Lee1 Minsoo Kang1 Bohyung Han1,2 1ECE & 2IPAI, Seoul National University {philip21, kminsoo, bhhan}@snu.ac.kr
Pseudocode Yes Algorithm 1 shows the detailed procedure of CSG.
Open Source Code Yes Code is available at https://github.com/Hleephilip/CSG.
Open Datasets Yes For quantitative evaluation, we select 250 images with the highest CLIP similarity from the LAION 5B dataset [20].
Dataset Splits No The paper mentions using a subset of the LAION 5B dataset for evaluation but does not specify explicit train/validation/test splits, percentages, or predefined partitions.
Hardware Specification Yes Our method is implemented based on the publicly available official code of Pix2Pix-Zero [35] in Py Torch [44] and tested on a single NVIDIA A100 GPU.
Software Dependencies No The paper mentions 'Py Torch' but does not specify a version number for it or any other software dependencies.
Experiment Setup Yes For faster generation, we adopt 50 forward steps using Eq. (1) and 50 reverse steps for target image generation. For translating target images, all methods employ the classifier-free guidance [46].