Reconstruction Bottlenecks in Object-Centric Generative Models

2020/07/13 by Martin Engelcke, Engelcke, Martin, Ōiwi Parker Jones +3
Computer Science · Mathematics · #Cellular Automata and Applications #Computability, Logic, AI Algorithms #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Markov Chains and Monte Carlo Methods

paper · pdf · doi:10.48550/arxiv.2007.06245

openalex publication_date 2020/07/13 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

A range of methods with suitable inductive biases exist to learn interpretable object-centric representations of images without supervision. However, these are largely restricted to visually simple images; robust object discovery in real-world sensory datasets remains elusive. To increase the understanding of such inductive biases, we empirically investigate the role of "reconstruction bottlenecks" for scene decomposition in GENESIS, a recent VAE-based model. We show such bottlenecks determine reconstruction and segmentation quality and critically influence model behaviour.

Citations

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Auto-Encoding Variational Bayes
• Efficient inference in occlusion-aware generative models of images
• Opening the Black Box of Deep Neural Networks via Information
• Relational Neural Expectation Maximization: Unsupervised Discovery of Objects and their Interactions
• Sequential Attend, Infer, Repeat: Generative Modelling of Moving Objects
• MONet: Unsupervised Scene Decomposition and Representation
• Spatial Broadcast Decoder: A Simple Architecture for Learning Disentangled Representations in VAEs
• COBRA: Data-Efficient Model-Based RL through Unsupervised Object Discovery and Curiosity-Driven Exploration
• Object Discovery with a Copy-Pasting GAN
• Emergence of Object Segmentation in Perturbed Generative Models
• SCALOR: Generative World Models with Scalable Object Representations
• Towards causal generative scene models via competition of experts

Related

• GENESIS: Generative Scene Inference and Sampling with Object-Centric Latent Representations
• Inductive Biases for Object-Centric Representations in the Presence of Complex Textures
• GENESIS-V2: Inferring Unordered Object Representations without Iterative Refinement
• Generative Models: What Do They Know? Do They Know Things? Let's Find Out!
• Uncovering Conceptual Blindspots in Generative Image Models Using Sparse Autoencoders