Stacked Capsule Autoencoders

Abstract

Objects are composed of a set of geometrically organized parts. We introduce an unsupervised capsule autoencoder (SCAE), which explicitly uses geometric relationships between parts to reason about objects. Since these relationships do not depend on the viewpoint, our model is robust to viewpoint changes. SCAE consists of two stages. In the first stage, the model predicts presences and poses of part templates directly from the image and tries to reconstruct the image by appropriately arranging the templates. In the second stage, SCAE predicts parameters of a few object capsules, which are then used to reconstruct part poses. Inference in this model is amortized and performed by off-the-shelf neural encoders, unlike in previous capsule networks. We find that object capsule presences are highly informative of the object class, which leads to state-of-the-art results for unsupervised classification on SVHN (55%) and MNIST (98.7%). The code is available at https://github.com/google-research/google-research/tree/master/stacked_capsule_autoencoders

arXiv commentNeurIPS 2019; 14 pages, 7 figures, 4 tables, code is available at https://github.com/google-research/google-research/tree/master/stacked_capsule_autoencoders

References in corpus (12)

• Learning deep representations by mutual information estimation and maximization
• The Concrete Distribution: A Continuous Relaxation of Discrete Random Variables
• Semi-Supervised Learning with Ladder Networks
• Group Equivariant Convolutional Networks
• Learning Discrete Representations via Information Maximizing Self-Augmented Training
• Multi-Object Representation Learning with Iterative Variational Inference
• MONet: Unsupervised Scene Decomposition and Representation
• GENESIS: Generative Scene Inference and Sampling with Object-Centric Latent Representations
• Group Equivariant Capsule Networks
• Sequential Attend, Infer, Repeat: Generative Modelling of Moving Objects
• Sparse Unsupervised Capsules Generalize Better
• Neural Network Encapsulation

Cited by in corpus (10)

• State estimation with limited sensors -- A deep learning based approach
• Graph Capsule Aggregation for Unaligned Multimodal Sequences
• SubTab: Subsetting Features of Tabular Data for Self-Supervised Representation Learning
• Faster Convergence in Deep-Predictive-Coding Networks to Learn Deeper Representations
• Capsule Network is Not More Robust than Convolutional Network
• Wasserstein Routed Capsule Networks
• Affine Self Convolution
• Capsule Networks -- A Probabilistic Perspective
• Deep Transformation-Invariant Clustering
• P-CapsNets: a General Form of Convolutional Neural Networks