mlsanityData to 7 Oct 2026

Stacked Capsule Autoencoders

arXiv 1906.06818 ↗First posted 17 Jun 2019stat.ML cs.CV cs.LGcited by 82

Not on the map. Why?

Abstract (arXiv)

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

Key facts. Cited by 82 papers in the map (data through 7 Oct 2026). First posted on arXiv 17 Jun 2019.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 32; 2022: 19; 2023: 19; 2024: 7; 2025: 1; 2026: 4.

82 papers.

Sort:

By area:

  1. 2021Interpretable Machine Learning: Fundamental Principles and 10 Grand Challengescited by 364
  2. 2021How to represent part-whole hierarchies in a neural networkcited by 123
  3. 2021Generative Adversarial Transformerscited by 120
  4. 2021GENESIS-V2: Inferring Unordered Object Representations without Iterative Refinementcited by 109
  5. 2021SubTab: Subsetting Features of Tabular Data for Self-Supervised Representation Learningcited by 101
  6. 2021Unsupervised Part Discovery from Contrastive Reconstructioncited by 57
  7. 2023Invariant Slot Attention: Object Discovery with Slot-Centric Reference Framescited by 51
  8. 2021Unsupervised Layered Image Decomposition into Object Prototypescited by 50
  9. 2022Unsupervised Learning of Group Invariant and Equivariant Representationscited by 38
  10. 2021GraphFPN: Graph Feature Pyramid Network for Object Detectioncited by 25
  11. 2021DECA: Deep viewpoint-Equivariant human pose estimation using Capsule Autoencoderscited by 23
  12. 2023Learning Layer-wise Equivariances Automatically using Gradientscited by 23
38 more
  1. 2023Towards Unsupervised Object Detection From LiDAR Point Cloudscited by 22
  2. 2021Visual Parser: Representing Part-whole Hierarchies with Transformerscited by 21
  3. 2021Capturing the objects of vision with neural networkscited by 21
  4. 2021Unsupervised Motion Representation Learning with Capsule Autoencoderscited by 16
  5. 2022Option-Aware Adversarial Inverse Reinforcement Learning for Robotic Controlcited by 16
  6. 2021Group Equivariant Subsamplingcited by 15
  7. 2023Multi-task Hierarchical Adversarial Inverse Reinforcement Learningcited by 15
  8. 2021Capsule Network is Not More Robust than Convolutional Networkcited by 12
  9. 2023Visual Dependency Transformers: Dependency Tree Emerges from Reversed Attentioncited by 12
  10. 2024Representing Part-Whole Hierarchies in Foundation Models by Learning Localizability, Composability, and Decomposability from Anatomy via Self-Supervisioncited by 12
  11. 2022Interpretable part-whole hierarchies and conceptual-semantic relationships in neural networkscited by 11
  12. 2021Reasons, Values, Stakeholders: A Philosophical Framework for Explainable Artificial Intelligencecited by 10
  13. 2022Learning with Capsules: A Surveycited by 10
  14. 2022Active Predictive Coding: A Unified Neural Framework for Learning Hierarchical World Models for Perception and Planningcited by 10
  15. 2023Are Deep Neural Networks Adequate Behavioural Models of Human Visual Perception?cited by 10
  16. 2022Test-time Adaptation with Slot-Centric Modelscited by 9
  17. 2021State estimation with limited sensors -- A deep learning based approachcited by 7
  18. 2022HP-Capsule: Unsupervised Face Part Discovery by Hierarchical Parsing Capsule Networkcited by 7
  19. 2021Graph Capsule Aggregation for Unaligned Multimodal Sequencescited by 6
  20. 2021Unsupervised Image Decomposition with Phase-Correlation Networkscited by 6
  21. 2022Active Predictive Coding Networks: A Neural Solution to the Problem of Learning Reference Frames and Part-Whole Hierarchiescited by 6
  22. 2023Why Capsule Neural Networks Do Not Scale: Challenging the Dynamic Parse-Tree Assumptioncited by 6
  23. 2021XnODR and XnIDR: Two Accurate and Fast Fully Connected Layers For Convolutional Neural Networkscited by 5
  24. 2021Routing with Self-Attention for Multimodal Capsule Networkscited by 5
  25. 2022Drawing out of Distribution with Neuro-Symbolic Generative Modelscited by 5
  26. 2023Learnable Earth Parser: Discovering 3D Prototypes in Aerial Scanscited by 5
  27. 2023Self-Supervised Detection of Perfect and Partial Input-Dependent Symmetriescited by 4
  28. 2024Unsupervised Object Discovery: A Comprehensive Survey and Unified Taxonomycited by 4
  29. 2021Deformable Capsules for Object Detectioncited by 3
  30. 2021Constellation: Learning relational abstractions over objects for compositional imaginationcited by 3
  31. 2022Holographic-(V)AE: an end-to-end SO(3)-Equivariant (Variational) Autoencoder in Fourier Spacecited by 3
  32. 2023Structured Generative Models for Scene Understandingcited by 3
  33. 2023Graphics Capsule: Learning Hierarchical 3D Face Representations from 2D Imagescited by 3
  34. 2021Faster Convergence in Deep-Predictive-Coding Networks to Learn Deeper Representationscited by 2
  35. 2021Inference for Generative Capsule Modelscited by 2
  36. 2021DPR-CAE: Capsule Autoencoder with Dynamic Part Representation for Image Parsingcited by 2
  37. 2021Neural Mixture Models with Expectation-Maximization for End-to-end Deep Clusteringcited by 2
  38. 2021Shifting Capsule Networks from the Cloud to the Deep Edgecited by 2

50 most cited of 82 All 82 (JSON)