Abstract:Feature pooling is a key component in modern visual classification system. However, the conventional two prevailing pooling techniques, namely average and max poolings, are not theoretically optimal, due to the unrecoverable loss of the spatial information during the statistical summarization and the underlying over-simplified assumption about the feature distribution. Addressing these issues, this paper proposes to generalize previous pooling methods toward a weighted ℓp-norm spatial pooling function tailored for class-specific feature spatial distribution. Optimizing such a pooling function toward discriminative class separability that is subject to a spatial smoothness constraint yields a so-called geometric ℓp-norm pooling (GLP) method. Furthermore, to handle the variation of object scale/position, which would affect not only the learning of discriminative pooling weights but also the applicability of the learned weights, we propose a simple yet effective self-alignment step during both learning and testing to adaptively adjust the pooling weights for individual images. Image segmentation and visual saliency map are utilized to construct a directed pixel adjacency graph. The discriminative pooling weights are diffused using random walk on the constructed graph and therefore the discriminative pooling weights are propagated onto the salient and foreground region. This leads to a robust version of GLP (RGLP) which can cope with the misalignment of object position and scale in images. Comprehensive experiments validate the effectiveness of the proposed GLP feature pooling framework. The proposed random walk based self-alignment step can effectively alleviate the image misalignment issue and further boost classification accuracy. State-of-the-art image classification and action recognition performances are attained on several benchmarks.

Geometric Neural Phrase Pooling: Modeling the Spatial Co-occurrence of Neurons

Spatial Pyramid Pooling Mechanism in 3D Convolutional Network for Sentence-Level Classification

Geometric ℓp-Norm Feature Pooling for Image Classification

SPGNN: Recognizing Salient Subgraph Patterns via Enhanced Graph Convolution and Pooling

Cross-convolutional-layer Pooling for Generic Visual Recognition.

Grassmann Pooling As Compact Homogeneous Bilinear Pooling For Fine-Grained Visual Classification

Deep CNNs Meet Global Covariance Pooling: Better Representation and Generalization

Geom-GCN: Geometric Graph Convolutional Networks

Generalized Pooling Pyramid with Hierarchical Dictionary Sparse Coding for Event and Object Recognition.

High-Order Local Pooling and Encoding Gaussians over a Dictionary of Gaussians.

Robust Geometric ℓp-Norm Feature Pooling for Image Classification and Action Recognition

Geometric L(P)-Norm Feature Pooling for Image Classification

Global Second-order Pooling Neural Networks

Generalizing Pooling Functions in Convolutional Neural Networks: Mixed, Gated, and Tree

Bag of Surrogate Parts Feature for Visual Recognition

Scene Understanding Based on Multi-Scale Pooling of Deep Learning Features

Pooling in Graph Convolutional Neural Networks

GOV-NeSF: Generalizable Open-Vocabulary Neural Semantic Fields

Spatial Group-wise Enhance: Improving Semantic Feature Learning in Convolutional Networks

An Improved Graph Pooling Network for Skeleton-Based Action Recognition

Convolutional Poisson Gamma Belief Network