Image Editing

The Image Shadow Triplets dataset (ISTD) is a dataset for shadow understanding that contains 1870 image triplets of shadow image, shadow mask, and shadow-free image. Source: ST-CGAN: "Stacked Conditional Generative Adversarial Networks for Jointly Learning Shadow Detection and Shadow Removal" Image Source: Stacked Conditional Generative Adversarial Networks for Jointly Learning Shadow Detection and Shadow Removal

SRD is a dataset for shadow removal that contains 3088 shadow and shadow-free image pairs.

ISTD+ consists of shadow images, shadow-free images, and shadow masks, with 1,330 training images and 540 testing images from 135 unique background scenes. ISTD suffers from color and luminosity inconsistencies between shadow and shadow-free images, which ISTD+ corrects with a color compensation mechanism to ensure uniform pixel colors across the ground-truth images.

ImgEdit is a large-scale, high-quality image-editing dataset comprising 1.2 million carefully curated edit pairs, which contain both novel and complex single-turn edits, as well as challenging multi-turn tasks.

Dataset for document shadow removal

BS-RSC is a real-world rolling shutter (RS) correction dataset and a corresponding model to correct the RS frames in a distorted video. Real distorted videos with corresponding ground truth are recorded simultaneously via a well-designed beam-splitter-based acquisition system. BSRSC contains various motions of both camera and objects in dynamic scenes.

Dataset for document shadow removal

This dataset is a new benchmark, grounded in real-world usages is developed to support more authentic and comprehensive evaluation of image editing models.

A significant challenge in removing shadows from indoor scenes is obtaining shadow-free images. To overcome this challenge, we propose a novel rendering pipeline for generating shadowed and shadow-free images under direct and indirect illumination, and create a comprehensive synthetic dataset that contains over 30,000 image pairs, covering various object types and lighting conditions. We implemented a direct/indirect shadow and shadow-free rendering pipeline using Blender Cycles engine, with the help of Open Shading Language (OSL). The resulting collection of shadow and shadow-free images is referred to as the “INS dataset”. The dataset includes 30,000 training and 2,000 testing images, all with a resolution of 512 × 512. The training and testing images are generated from distinct scenes with different objects and materials.

Doc3DShade extends Doc3D with realistic lighting and shading. Follows a similar synthetic rendering procedure using captured document 3D shapes but final image generation step combines real shading of different types of paper materials under numerous illumination conditions. Source: Doc3DShade

SD7K is the only large-scale high-resolution dataset that satisfies all important data features about document shadow currently, which covers a large number of document shadow images. Mean resolution is $2462 \times 3699$

A version of the WSRD Dataset will be used as a benchmark for the NTIRE24 Challenge on Image Shadow Removal.