Unsupervised Object Segmentation by Redrawing

Abstract

Object segmentation is a crucial problem that is usually solved by using supervised learning approaches over very large datasets composed of both images and corresponding object masks. Since the masks have to be provided at pixel level, building such a dataset for any new domain can be very time-consuming. We present ReDO, a new model able to extract objects from images without any annotation in an unsupervised way. It relies on the idea that it should be possible to change the textures or colors of the objects without changing the overall distribution of the dataset. Following this assumption, our approach is based on an adversarial architecture where the generator is guided by an input sample: given an image, it extracts the object mask, then redraws a new object at the same location. The generator is controlled by a discriminator that ensures that the distribution of generated images is aligned to the original one. We experiment with this method on different datasets and demonstrate the good quality of extracted masks.

arXiv commentPresented at NeurIPS 2019

References in corpus (20)

• Adam: A Method for Stochastic Optimization
• Large Scale GAN Training for High Fidelity Natural Image Synthesis
• Spectral Normalization for Generative Adversarial Networks
• Self-Attention Generative Adversarial Networks
• Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
• Conditional Adversarial Domain Adaptation
• A Learned Representation For Artistic Style
• Disentangling by Factorising
• Augmented CycleGAN: Learning Many-to-Many Mappings from Unpaired Data
• Disentangling factors of variation in deep representations using adversarial training
• Multi-Object Representation Learning with Iterative Variational Inference
• W-Net: A Deep Model for Fully Unsupervised Image Segmentation
• LR-GAN: Layered Recursive Generative Adversarial Networks for Image Generation
• MONet: Unsupervised Scene Decomposition and Representation
• Structured Disentangled Representations
• Hierarchical Implicit Models and Likelihood-Free Variational Inference
• High-Fidelity Image Generation With Fewer Labels
• On Self Modulation for Generative Adversarial Networks
• Investigating Object Compositionality in Generative Adversarial Networks
• SEIGAN: Towards Compositional Image Generation by Simultaneously Learning to Segment, Enhance, and Inpaint

Cited by in corpus (7)

• Investigating Object Compositionality in Generative Adversarial Networks
• Coherent, super resolved radar beamforming using self-supervised learning
• GENESIS-V2: Inferring Unordered Object Representations without Iterative Refinement
• Unsupervised Medical Image Segmentation with Adversarial Networks: From Edge Diagrams to Segmentation Maps
• Reconstruction Bottlenecks in Object-Centric Generative Models
• Unsupervised Object Segmentation with Explicit Localization Module
• Polyline Generative Navigable Space Segmentation for Autonomous Visual Navigation