mlsanityData to 7 Oct 2026

Interpretable Compositional Convolutional Neural Networks

arXiv 2107.04474 ↗First posted 9 Jul 2021cs.CVcited by 15

Not on the map. Why?

Abstract (arXiv)

The reasonable definition of semantic interpretability presents the core challenge in explainable AI. This paper proposes a method to modify a traditional convolutional neural network (CNN) into an interpretable compositional CNN, in order to learn filters that encode meaningful visual patterns in intermediate convolutional layers. In a compositional CNN, each filter is supposed to consistently represent a specific compositional object part or image region with a clear meaning. The compositional CNN learns from image labels for classification without any annotations of parts or regions for supervision. Our method can be broadly applied to different types of CNNs. Experiments have demonstrated the effectiveness of our method.

Key facts. Cited by 15 papers in the map (data through 7 Oct 2026). Builds on 17 of its 32 references that are in the map. First posted on arXiv 9 Jul 2021.

Builds on

17 of 32 references.

References · 17

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,267
  3. 2015Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shiftcited by 6,979
  4. 2016Densely Connected Convolutional Networkscited by 6,691
  5. 2014Deep Learning Face Attributes in the Wildcited by 3,679
  6. 2016InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Netscited by 858
  7. 2018This Looks Like That: Deep Learning for Interpretable Image Recognitioncited by 731
  8. 2017Network Dissection: Quantifying Interpretability of Deep Visual Representationscited by 703
  9. 2017Dynamic Routing Between Capsulescited by 689
  10. 2014Detect What You Can: Detecting and Representing Objects using Holistic Models and Body Partscited by 230
  11. 2017Interpretable Convolutional Neural Networkscited by 161
  12. 2018Net2Vec: Quantifying and Explaining how Concepts are Encoded by Filters in Deep Neural Networkscited by 128
5 more
  1. 2020Interpretable and Accurate Fine-grained Recognition via Region Groupingcited by 75
  2. 2020Compositional Convolutional Neural Networks: A Deep Architecture with Innate Robustness to Partial Occlusioncited by 47
  3. 2020Training Interpretable Convolutional Neural Networks by Differentiating Class-specific Filterscited by 26
  4. 2014Semantic Part Segmentation using Compositional Model combining Shape and Appearancecited by 22
  5. 2017Teaching Compositionality to CNNscited by 18

Built on by

Cited by year

Citing papers per year: 2022: 4; 2023: 3; 2024: 4; 2025: 4.

15 papers.

Sort:

  1. 2022Quantifying the Knowledge in a DNN to Explain Knowledge Distillation for Classificationcited by 19
  2. 2025A Comprehensive Survey on Self-Interpretable Neural Networkscited by 18
  3. 2022ME-D2N: Multi-Expert Domain Decompositional Network for Cross-Domain Few-Shot Learningcited by 9
  4. 2023Adversarial Attacks on the Interpretation of Neuron Activation Maximizationcited by 7
  5. 2024A Large-Scale Empirical Study on Improving the Fairness of Image Classification Modelscited by 6
  6. 2022Bort: Towards Explainable Neural Networks with Bounded Orthogonal Constraintcited by 5
  7. 2024Path Choice Matters for Clear Attribution in Path Methodscited by 4
  8. 2024Layerwise Change of Knowledge in Neural Networkscited by 4
  9. 2023PICNN: A Pathway towards Interpretable Convolutional Neural Networkscited by 2
  10. 2024Improving Network Interpretability via Explanation Consistency Evaluationcited by 2
  11. 2025A Survey on Interpretability in Visual Recognitioncited by 2
  12. 2025Disentangling Visual Transformers: Patch-level Interpretability for Image Classificationcited by 1
3 more
  1. 2022TCNL: Transparent and Controllable Network Learning Via Embedding Human-Guided Conceptscited by 0
  2. 2023DBAT: Dynamic Backward Attention Transformer for Material Segmentation with Cross-Resolution Patchescited by 0
  3. 2025Concept Labels Are Not Enough: Rethinking Concept Bottleneck Models through Representation Integritycited by 0