Video Captioning With Attention-Based LSTM and Semantic Consistency

TLDR

Recent advances in LSTM‑based image captioning have spurred interest in applying them to video captioning, yet most approaches compress entire shots into static representations, neglecting attention mechanisms and the semantic alignment between visual content and generated sentences. The authors introduce aLSTMs, an end‑to‑end attention‑based LSTM framework with semantic consistency, to generate natural sentences from videos. aLSTMs integrates an attention module that dynamically weights local CNN features, an LSTM decoder that consumes these visual cues and previous word embeddings, and a multimodal embedding that aligns visual and textual representations in a joint space to enforce semantic consistency. On benchmark datasets, aLSTMs achieves competitive or superior BLEU and METEOR scores compared to state‑of‑the‑art video captioning baselines using only a single feature.

Abstract

Recent progress in using long short-term memory (LSTM) for image captioning has motivated the exploration of their applications for video captioning. By taking a video as a sequence of features, an LSTM model is trained on video-sentence pairs and learns to associate a video to a sentence. However, most existing methods compress an entire video shot or frame into a static representation, without considering attention mechanism which allows for selecting salient features. Furthermore, existing approaches usually model the translating error, but ignore the correlations between sentence semantics and visual content. To tackle these issues, we propose a novel end-to-end framework named aLSTMs, an attention-based LSTM model with semantic consistency, to transfer videos to natural sentences. This framework integrates attention mechanism with LSTM to capture salient structures of video, and explores the correlation between multimodal representations (i.e., words and visual content) for generating sentences with rich semantic content. Specifically, we first propose an attention mechanism that uses the dynamic weighted sum of local two-dimensional convolutional neural network representations. Then, an LSTM decoder takes these visual features at time t and the word-embedding feature at time t-1 to generate important words. Finally, we use multimodal embedding to map the visual and sentence features into a joint space to guarantee the semantic consistence of the sentence description and the video visual content. Experiments on the benchmark datasets demonstrate that our method using single feature can achieve competitive or even better results than the state-of-the-art baselines for video captioning in both BLEU and METEOR.