Object Discovery with a Copy-Pasting GAN

Abstract

We tackle the problem of object discovery, where objects are segmented for a given input image, and the system is trained without using any direct supervision whatsoever. A novel copy-pasting GAN framework is proposed, where the generator learns to discover an object in one image by compositing it into another image such that the discriminator cannot tell that the resulting image is fake. After carefully addressing subtle issues, such as preventing the generator from `cheating', this game results in the generator learning to select objects, as copy-pasting objects is most likely to fool the discriminator. The system is shown to work well on four very different datasets, including large object appearance variations in challenging cluttered backgrounds.

References in corpus (16)

• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
• Learning to Segment Object Candidates
• Adversarial Feature Learning
• Associative Embedding: End-to-End Learning for Joint Detection and Grouping
• Semantic Instance Segmentation with a Discriminative Loss Function
• Image Splicing Localization Using A Multi-Task Fully Convolutional Network (MFCN)
• Attend, Infer, Repeat: Fast Scene Understanding with Generative Models
• Multi-Object Representation Learning with Iterative Variational Inference
• Semantic Instance Segmentation via Deep Metric Learning
• MONet: Unsupervised Scene Decomposition and Representation
• Synthesizing Programs for Images using Reinforced Adversarial Learning
• Patch-Based Image Inpainting with Generative Adversarial Networks
• Context-Aware Synthesis and Placement of Object Instances
• Unsupervised Cipher Cracking Using Discrete GANs
• TF-Replicator: Distributed Machine Learning for Researchers
• Learning Dense Convolutional Embeddings for Semantic Segmentation

Cited by in corpus (15)

• GENESIS: Generative Scene Inference and Sampling with Object-Centric Latent Representations
• On the Binding Problem in Artificial Neural Networks
• Investigating Object Compositionality in Generative Adversarial Networks
• Finding an Unsupervised Image Segmenter in Each of Your Deep Generative Models
• Object-aware Contrastive Learning for Debiased Scene Representation
• GENESIS-V2: Inferring Unordered Object Representations without Iterative Refinement
• Compositional Transformers for Scene Generation
• Distilling Localization for Self-Supervised Representation Learning
• Context Decoupling Augmentation for Weakly Supervised Semantic Segmentation
• Multiple-Identity Image Attacks Against Face-based Identity Verification
• Reconstruction Bottlenecks in Object-Centric Generative Models
• Learning to Manipulate Individual Objects in an Image
• Self-supervised Segmentation via Background Inpainting
• Conditional Adversarial Camera Model Anonymization
• Disassembling Object Representations without Labels