Learning Contextual Dependencies with Convolutional Hierarchical Recurrent Neural Networks

TL;DR

This paper introduces convolutional hierarchical recurrent neural networks (C-HRNNs) that combine CNNs with hierarchical RNNs to effectively model spatial and scale dependencies in images, leading to state-of-the-art classification results.

Contribution

The paper proposes a novel integration of CNNs with hierarchical RNNs, specifically HSRN and HLSTM models, to encode contextual dependencies in images for improved classification.

Findings

Achieved state-of-the-art results on Places 205 and SUN 397 datasets.

Demonstrated effective modeling of spatial and scale dependencies.

Presented two RNN models with different computational costs and performance.

Abstract

Existing deep convolutional neural networks (CNNs) have shown their great success on image classification. CNNs mainly consist of convolutional and pooling layers, both of which are performed on local image areas without considering the dependencies among different image regions. However, such dependencies are very important for generating explicit image representation. In contrast, recurrent neural networks (RNNs) are well known for their ability of encoding contextual information among sequential data, and they only require a limited number of network parameters. General RNNs can hardly be directly applied on non-sequential data. Thus, we proposed the hierarchical RNNs (HRNNs). In HRNNs, each RNN layer focuses on modeling spatial dependencies among image regions from the same scale but different locations. While the cross RNN scale connections target on modeling scale dependencies among regions from the same location but different scales. Specifically, we propose two recurrent neural network models: 1) hierarchical simple recurrent network (HSRN), which is fast and has low computational cost; and 2) hierarchical long-short term memory recurrent network (HLSTM), which performs better than HSRN with the price of more computational cost. In this manuscript, we integrate CNNs with HRNNs, and develop end-to-end convolutional hierarchical recurrent neural networks (C-HRNNs). C-HRNNs not only make use of the representation power of CNNs, but also efficiently encodes spatial and scale dependencies among different image regions. On four of the most challenging object/scene image classification benchmarks, our C-HRNNs achieve state-of-the-art results on Places 205, SUN 397, MIT indoor, and competitive results on ILSVRC 2012.