Do Semantic Parts Emerge in Convolutional Neural Networks?

By Abel Gonzalez-Garcia, Davide Modolo, Vittorio Ferrari. 2016

118 citations recorded by the source corpus.

Semantic object parts can be useful for several visual recognition tasks. Lately, these tasks have been addressed using Convolutional Neural Networks (CNN), achieving outstanding results. In this work we study whether CNNs learn semantic parts in their internal representation. We investigate the responses of convolutional filters and try to associate their stimuli with semantic parts. We perform two extensive quantitative analyses. First, we use ground-truth part bounding-boxes from the PASCAL-Part dataset to determine how many of those semantic parts emerge in the CNN. We explore this emergence for different layers, network depths, and supervision levels. Second, we collect human judgements in order to study what fraction of all filters systematically fire on any semantic part, even if not annotated in PASCAL-Part. Moreover, we explore several connections between discriminative power and semantics. We find out which are the most discriminative filters for object recognition, and analyze whether they respond to semantic parts or to other image patches. We also investigate the other direction: we determine which semantic parts are the most discriminative and whether they correspond to those parts emerging in the network. This enables to gain an even deeper understanding of the role of semantic parts in the network.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Representation & Self-Supervised Learning, Model Architectures

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Representation & Self-Supervised Learning

In Model Architectures

Citations in the corpus

Network Dissection: Quantifying Interpretability of Deep Visual Representations (2017, background): “Observations of hidden units in large deep neural networks have revealed that human-interpretable concepts sometimes emerge as individual latent variables within those networks: for example, object detector units emerge within networks trained to recognize places [40]; part detectors emerge in objec”
Interpretable Basis Decomposition for Visual Explanation (2018, background): “A second approach is to identify the purpose of the internal representations of a network by identifying the concepts that each part of the network detects [3, 27, 7].”
Interpreting Deep Visual Representations via Network Dissection (2017, background): “For example, object detector units emerge within networks trained to recognize places [1], part detectors emerge in object classifiers [2] and object detectors emerge in generative video networks [3].”
Counterfactual Visual Explanations (2019, positive): “This is in agreement with prior work (Zhou et al., 2015; Gonzalez-Garcia et al., 2018) that show emergence of semantic detectors in the internal representations of classification networks.”
Occluded Pedestrian Detection Through Guided Attention in CNNs (2018, background): “For instance, [1] performs network dissection and finds that many individual units respond to specific high-level concepts and [28, 10] also find some filter responses can be linked to semantic parts.”
Deep Feature Factorization For Concept Discovery (2018, methodology): “Unlike previous approaches [2, 11], we do not require annotated data to detect semantic features.”