데이터셋 탐색

The DreamBooth dataset is a collection of images used for fine-tuning text-to-image diffusion models for subject-driven generation¹. Here are some key details about the dataset: - The dataset includes 30 subjects from 15 different classes¹. - Among these subjects, 9 are live subjects (such as dogs and cats) and 21 are objects¹. - The dataset contains a variable number of images per subject, typically between 4 to 6 images¹. - Images of the subjects are usually captured in different conditions, environments, and under different angles¹. - The dataset also includes a file prompts_and_classes.txt which contains all of the prompts used in the paper for live subjects and objects, as well as the class name used for the subjects¹. - The images have either been captured by the paper authors or sourced from www.unsplash.com¹. - The references_and_licenses.txt file contains a list of all the reference links to the images in www.unsplash.com, along with the attribution to the photographer and the license of the image¹. This dataset is part of the official repository for the Google paper "DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation"¹. If you use this work, please cite the paper¹. Please note that this is not an officially supported Google product¹. (1) GitHub - google/dreambooth. https://github.com/google/dreambooth. (2) DreamBooth - Hugging Face. https://huggingface.co/docs/diffusers/training/dreambooth. (3) google/dreambooth · Datasets at Hugging Face. https://huggingface.co/datasets/google/dreambooth. (4) dreambooth: Mirror of https://huggingface.co/datasets/google .... https://gitee.com/hf-datasets/dreambooth. (5) undefined. https://github.com/huggingface/diffusers. (6) undefined. https://huggingface.co/datasets/google.

DreamBench++ is a human-aligned benchmark for personalized image generation¹. It's automated by advanced multimodal GPT models¹. The goal of DreamBench++ is to assist humans in everyday work and life by creatively generating personalized content¹². Here are some key features of DreamBench++: - Prompts Design: The prompts are systematically designed to let GPT be both human-aligned and self-aligned¹. - Task Reinforcement: The GPT models are empowered with task reinforcement¹. - Diverse Dataset: A comprehensive dataset comprising diverse images and prompts is constructed¹. - Human-Aligned Evaluation: By benchmarking 7 modern generative models, DreamBench++ results in significantly more human-aligned evaluation¹. (1) DreamBench++: A Human-Aligned Benchmark for Personalized Image Generation. https://dreambenchplus.github.io/. (2) DreamBench++: A Human-Aligned Benchmark for Personalized Image Generation. https://arxiv.org/abs/2406.16855. (3) 图像生成评估基准DREAMBENCH++：与人类偏好对齐的 .... https://sd114.wiki/2676.html. (4) DreamBench++: A Human-Aligned Benchmark for Personalized Image Generation. https://arxiv.org/html/2406.16855v1.