# SEIGAN: Towards Compositional Image Generation by Simultaneously Learning to Segment, Enhance, and Inpaint

arXiv: https://arxiv.org/abs/1811.07630 · HTML: https://mlsanity.com/paper/1811.07630/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 19 Nov 2018.

## Built on by (most cited first, 9 of 9)

- [Unsupervised Foreground Extraction via Deep Region Competition](https://mlsanity.com/paper/2110.15497/) (2021)
- [MOVE: Unsupervised Movable Object Segmentation and Detection](https://mlsanity.com/paper/2210.07920/) (2022)
- [Unsupervised Object Localization in the Era of Self-Supervised ViTs: A Survey](https://mlsanity.com/paper/2310.12904/) (2023)
- [Behind the leaves -- Estimation of occluded grapevine berries with conditional generative adversarial networks](https://mlsanity.com/paper/2105.10325/) (2021)
- [GaussiGAN: Controllable Image Synthesis with 3D Gaussians from Unposed Silhouettes](https://mlsanity.com/paper/2106.13215/) (2021)
- [Edge-competing Pathological Liver Vessel Segmentation with Limited Labels](https://mlsanity.com/paper/2108.00384/) (2021)
- [Visual Boundary Knowledge Translation for Foreground Segmentation](https://mlsanity.com/paper/2108.00379/) (2021)
- [TricycleGAN: Unsupervised Image Synthesis and Segmentation Based on Shape Priors](https://mlsanity.com/paper/2102.02690/) (2021)
- [Weakly Supervised Object Segmentation by Background Conditional Divergence](https://mlsanity.com/paper/2506.22505/) (2025)

## Abstract (arXiv)

We present a novel approach to image manipulation and understanding by simultaneously learning to segment object masks, paste objects to another background image, and remove them from original images. For this purpose, we develop a novel generative model for compositional image generation, SEIGAN (Segment-Enhance-Inpaint Generative Adversarial Network), which learns these three operations together in an adversarial architecture with additional cycle consistency losses. To train, SEIGAN needs only bounding box supervision and does not require pairing or ground truth masks. SEIGAN produces better generated images (evaluated by human assessors) than other approaches and produces high-quality segmentation masks, improving over other adversarially trained approaches and getting closer to the results of fully supervised training.