Part Detector Discovery in Deep Convolutional Neural Networks

By Marcel Simon, Erik Rodner, Joachim Denzler. 2014

76 citations recorded by the source corpus.

Current fine-grained classification approaches often rely on a robust localization of object parts to extract localized feature representations suitable for discrimination. However, part localization is a challenging task due to the large variation of appearance and pose. In this paper, we show how pre-trained convolutional neural networks can be used for robust and efficient object part discovery and localization without the necessity to actually train the network on the current dataset. Our approach called “part detector discovery” (PDD) is based on analyzing the gradient maps of the network outputs and finding activation centers spatially related to annotated semantic parts or bounding boxes. This allows us not just to obtain excellent performance on the CUB200-2011 dataset, but in contrast to previous approaches also to perform detection and bird classification jointly without requiring a given bounding box annotation during testing and ground-truth parts during training.

Subfields: Computer Vision and Pattern Recognition, Model Architectures, Safety, Robustness & Interpretability

How researchers see this paper

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

In Computer Vision and Pattern Recognition

In Model Architectures

In Safety, Robustness & Interpretability

Citations in the corpus

Interpreting CNNs via Decision Trees (2018, methodology): “Like in [3, 28], we ignored species labels and regarded all these images as a single bird category.”
Interpretable Convolutional Neural Networks (2017, background): “discovered objects from feature maps of unlabeled images [22], and selected a certain filter to describe each semantic part in a supervised fashion [23].”
The Unreasonable Effectiveness of Noisy Data for Fine-Grained Recognition (2015, neutral): “The use of existing test sets serves to provide an accurate measure of performance and put our work in a larger context, but results may not be strictly comparable with prior work that operates within a single given dataset.”
Neural Activation Constellations: Unsupervised Part Model Discovery with Convolutional Networks (2015, methodology): “[29] presents a method, which requires bounding boxes of the object during training rather than part annotations.”
Interpreting CNN knowledge via an Explanatory Graph (2017, methodology): “Visualization of CNNs: Visualization of ﬁlters in a CNN is the most direct way of exploring the pattern hidden inside a neural unit.”
Do Semantic Parts Emerge in Convolutional Neural Networks? (2016, background): “Several works use filter responses for recognition tasks (Simon et al. 2014; Gkioxari et al. 2015; Simon and Rodner 2015; Xiao et al. 2015; Oquab et al. 2015).”
Low-Rank Bilinear Pooling for Fine-Grained Classification (2016, background): “One approach to unsupervised part discovery [27, 26] uses saliency maps, leveraging the observation that sparse deep CNN feature activations often correspond to semantically meaningful regions [34, 20].”
Occluded Pedestrian Detection Through Guided Attention in CNNs (2018, result): “Similar findings are shown in [25], that in a bird classification network some channels are associated with parts.”
Mining Object Parts from CNNs via Active Question-Answering (2017, methodology): “Like in [4, 32, 46], we ignored species labels and regarded all these images as a single bird category.”
Interactively Transferring CNN Patterns for Part Localization (2017, methodology): “∂p is equal to the gradient of pattern score w.r.t. pixel p (see [13] for details of the gradient computation).”
Cross-Dimensional Weighting for Aggregated Deep Convolutional Features (2015, background): “[30] start from a pre-trained CNN and discover part detectors in a supervised as well as unsupervised way.”
Making Sense of CNNs: Interpreting Deep Representations & Their Invariances with INNs (2020, background): “Different works attempt to select groups of neurons which have a certain semantic meaning, such as based on scenes [66], objects [55] and object parts [56].”
Fine-grained Recognition in the Noisy Wild: Sensitivity Analysis of Convolutional Neural Networks Approaches (2016, methodology): “The gradients can be easily computed with a backward pass [25, 27].”