Image Captioning with Deep Bidirectional LSTMs

TL;DR

This paper introduces a deep bidirectional LSTM model for image captioning that effectively captures long-term visual-language interactions, achieving competitive results on standard benchmarks without additional mechanisms.

Contribution

The paper proposes novel deep bidirectional LSTM architectures and data augmentation techniques for improved image captioning and retrieval performance.

Findings

Achieves state-of-the-art results on caption generation tasks.

Significantly outperforms recent methods on image-sentence retrieval.

Demonstrates the effectiveness of bidirectional LSTMs in modeling visual-language relationships.

Abstract

This work presents an end-to-end trainable deep bidirectional LSTM (Long-Short Term Memory) model for image captioning. Our model builds on a deep convolutional neural network (CNN) and two separate LSTM networks. It is capable of learning long term visual-language interactions by making use of history and future context information at high level semantic space. Two novel deep bidirectional variant models, in which we increase the depth of nonlinearity transition in different way, are proposed to learn hierarchical visual-language embeddings. Data augmentation techniques such as multi-crop, multi-scale and vertical mirror are proposed to prevent overfitting in training deep models. We visualize the evolution of bidirectional LSTM internal states over time and qualitatively analyze how our models "translate" image to sentence. Our proposed models are evaluated on caption generation and image-sentence retrieval tasks with three benchmark datasets: Flickr8K, Flickr30K and MSCOCO datasets. We demonstrate that bidirectional LSTM models achieve highly competitive performance to the state-of-the-art results on caption generation even without integrating additional mechanism (e.g. object detection, attention model etc.) and significantly outperform recent methods on retrieval task.