DESOBAv2: Towards Large-scale Real-world Dataset for Shadow Generation

Qingyang Liu, Jianting Wang, Li Niu
8/19/2023
cs.CV

Official code: bcmi/Object-Shadow-Generation-Dataset-DESOBAv2 (163 stars, Apache-2.0, last commit 2026-02-24).

Abstract

Image composition refers to inserting a foreground object into a background image to obtain a composite image. In this work, we focus on generating plausible shadow for the inserted foreground object to make the composite image more realistic. To supplement the existing small-scale dataset DESOBA, we create a large-scale dataset called DESOBAv2 by using object-shadow detection and inpainting techniques. Specifically, we collect a large number of outdoor scene images with object-shadow pairs. Then, we use pretrained inpainting model to inpaint the shadow region, resulting in the deshadowed images. Based on real images and deshadowed images, we can construct pairs of synthetic composite images and ground-truth target images. Dataset is available at https://github.com/bcmi/Object-Shadow-Generation-Dataset-DESOBAv2.

View on arXivView PDF

Code Implementations(1)

bcmi/Object-Shadow-Generation-Dataset-DESOBAv2Official100%
Apache-2.0

[CVPR 2024] The dataset, code, and model for our paper "Shadow Generation for Composite Image Using Diffusion Model", CVPR, 2024.

16314PythonAug 19, 20237 months agoApache-2.0
image-compositionshadow-generation

Similar Papers

Shadow Generation with Decomposed Mask Prediction and Attentive Shadow Filling

88% match1 repo222023

Shadow Generation for Composite Image Using Diffusion Model

85% match1 repo1632024

Shadow Generation Using Diffusion Model with Geometry Prior

84% match1 repo1522025

Controllable Shadow Generation with Single-Step Diffusion Models from Synthetic Data

82% match2024

ShadowDraw: From Any Object to Shadow-Drawing Compositional Art

82% match1 repo282025

Generative Portrait Shadow Removal

81% match2024

Cite this paper

@article{liu2023desobav,
  title  = {DESOBAv2: Towards Large-scale Real-world Dataset for Shadow Generation},
  author = {Qingyang Liu and Jianting Wang and Li Niu},
  year   = {2023},
  eprint = {2308.09972},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2308.09972v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2308.09972.svg)](https://codeofpaper.com/paper/2308.09972)

Discussion