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.

Multi-scale Gem Pooling with N-Pair Center Loss for Fine-Grained Image Search

Multi-View Graph Embedding Learning for Image Co-Segmentation and Co-Localization

Learning Feature Embedding with Strong Neural Activations for Fine-Grained Retrieval

Multi-FusNet: fusion mapping of features for fine-grained image retrieval networks

Selective Convolutional Features based Generalized-mean Pooling for Fine-grained Image Retrieval

Learning Representation of Multi-Scale Object for Fine-Grained Image Retrieval.

Improved fine-grained object retrieval with Hard Global Softmin Loss objective

Matryoshka Peek: Toward Learning Fine-Grained, Robust, Discriminative Features for Product Search

Rethinking Pooling for Multi-Granularity Features in Aerial-View Geo-Localization

Two-stage Pooling of Deep Convolutional Features for Image Retrieval.

Multi-scale Orderless Cross-Regions-pooling of Deep Attributes for Image Retrieval

REMAP: Multi-Layer Entropy-Guided Pooling of Dense CNN Features for Image Retrieval

Deep embedding learning with adaptive large margin N-pair loss for image retrieval and clustering

Coarse-to-Fine: Learning Compact Discriminative Representation for Single-Stage Image Retrieval

Multi-Index Fusion Via Similarity Matrix Pooling for Image Retrieval

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

A Lightweight Multi-Grained Image-Text Retrieval Paradigm Via Cascaded Representation Learning and Parameter-Free Feature Aggregation

Combining Local and Global: Rich and Robust Feature Pooling for Visual Recognition.

Fine-grained Visual Classification Via Multilayer Bilinear Pooling with Object Localization

A feature consistency driven attention erasing network for fine-grained image retrieval

Multi-modal Fine-grained Retrieval with Local and Global Cross-Attention.