26 datasets found

A dataset of images used to evaluate the proposed method for reducing visual distraction in images.

SemanticGAN uses a dataset of real images and their corresponding semantic segmentation masks.

DatasetGAN uses a dataset of real images and their corresponding semantic segmentation masks.

The dataset used in the paper Harnessing Human Feedback for Instructional Visual Editing (HIVE) for instructional visual editing.

A benchmark dataset for text-to-image diffusion models.

The MAGIC dataset is a multiclass classification dataset containing 19020 samples, each described by 10 features.

The dataset used for training the InstructAny2Pix model, containing 500k instructions with multi-modal inputs.

A dataset comprising 3000 images for quantitative evaluation of specified region customization with both text and image inputs.

Given a short monocular RGB video and text instructions, our method uses an image-conditioned diffusion model to edit one head image and uses the video stylization method to...

The Region-Aware Diffusion for Zero-Shot Text-Driven Image Editing dataset is used to evaluate the ability of models to edit images based on text descriptions.

The dataset used in the paper is the Image Multi-Adjustment Dataset (I-MAD-Dense). It contains approximately 22,000 triplets, each containing a source image, an edited image,...

The dataset used in the paper for text-driven image editing with diffusion models.

PIE-Bench is a dataset for single-object editing, featuring a modified subset of the dataset with images from AFHQ, COCO, and Imagen.

LoMOE-Bench is a dataset for multi-object editing, featuring 64 images with 2 to 7 masks, paired with corresponding text prompts.

Learning to follow image editing instructions.

The dataset used in the paper for text-driven image editing, consisting of synthetic images generated by Midjourney.

The dataset used in the paper for text-driven image editing, consisting of synthetic images generated by Midjourney and COCO dataset images.

The dataset used for training and testing the Conditionally-Independent Pixel Synthesis (CIPS) generator.

Flickr-Faces-HQ contains 70,000 face images at 1024 × 1024 resolution, which were originally crawled from Flickr, manually checked to discard low-quality samples, and then...

The dataset used in the paper is a set of images from the AFHQ dataset, containing 1.5K images of different animal faces.