Image Question Answering using Convolutional Neural Network with Dynamic Parameter Prediction

TL;DR

This paper introduces a CNN-based image question answering model with a dynamic parameter layer whose weights are adaptively predicted from questions using a GRU, employing hashing to manage complexity, achieving state-of-the-art results.

Contribution

It proposes a novel dynamic parameter prediction mechanism for CNNs in ImageQA, utilizing hashing to efficiently handle large parameter spaces, and demonstrates superior performance.

Findings

Achieves state-of-the-art results on public ImageQA benchmarks.

Introduces a hashing technique to reduce complexity in dynamic parameter prediction.

End-to-end training of the joint network enhances performance.

Abstract

We tackle image question answering (ImageQA) problem by learning a convolutional neural network (CNN) with a dynamic parameter layer whose weights are determined adaptively based on questions. For the adaptive parameter prediction, we employ a separate parameter prediction network, which consists of gated recurrent unit (GRU) taking a question as its input and a fully-connected layer generating a set of candidate weights as its output. However, it is challenging to construct a parameter prediction network for a large number of parameters in the fully-connected dynamic parameter layer of the CNN. We reduce the complexity of this problem by incorporating a hashing technique, where the candidate weights given by the parameter prediction network are selected using a predefined hash function to determine individual weights in the dynamic parameter layer. The proposed network---joint network with the CNN for ImageQA and the parameter prediction network---is trained end-to-end through back-propagation, where its weights are initialized using a pre-trained CNN and GRU. The proposed algorithm illustrates the state-of-the-art performance on all available public ImageQA benchmarks.