The MSR-Video to Text Dataset with Clean Annotations

TL;DR

This paper improves video captioning by cleaning the MSR-VTT dataset annotations, leading to better model performance and more coherent, relevant captions according to human evaluation.

Contribution

The authors identified noise in MSR-VTT annotations, cleaned the dataset, and demonstrated improved model performance and caption quality after cleaning.

Findings

Data cleaning increased model performance metrics.

Humans preferred captions from the cleaned dataset.

Cleaned dataset produced more coherent and relevant captions.

Abstract

Video captioning automatically generates short descriptions of the video content, usually in form of a single sentence. Many methods have been proposed for solving this task. A large dataset called MSR Video to Text (MSR-VTT) is often used as the benchmark dataset for testing the performance of the methods. However, we found that the human annotations, i.e., the descriptions of video contents in the dataset are quite noisy, e.g., there are many duplicate captions and many captions contain grammatical problems. These problems may pose difficulties to video captioning models for learning underlying patterns. We cleaned the MSR-VTT annotations by removing these problems, then tested several typical video captioning models on the cleaned dataset. Experimental results showed that data cleaning boosted the performances of the models measured by popular quantitative metrics. We recruited subjects to evaluate the results of a model trained on the original and cleaned datasets. The human behavior experiment demonstrated that trained on the cleaned dataset, the model generated captions that were more coherent and more relevant to the contents of the video clips.