GRIT: Faster and Better Image captioning Transformer Using Dual Visual Features

TL;DR

GRIT is a Transformer-based image captioning model that combines grid and region features, replacing CNN detectors with DETR for faster, more accurate caption generation through end-to-end training.

Contribution

Introduces GRIT, a novel Transformer-only architecture that effectively fuses grid and region visual features, replacing CNN detectors with DETR for improved speed and accuracy.

Findings

Outperforms previous methods in captioning accuracy.

Achieves faster inference times.

Enables end-to-end training of the model.

Abstract

Current state-of-the-art methods for image captioning employ region-based features, as they provide object-level information that is essential to describe the content of images; they are usually extracted by an object detector such as Faster R-CNN. However, they have several issues, such as lack of contextual information, the risk of inaccurate detection, and the high computational cost. The first two could be resolved by additionally using grid-based features. However, how to extract and fuse these two types of features is uncharted. This paper proposes a Transformer-only neural architecture, dubbed GRIT (Grid- and Region-based Image captioning Transformer), that effectively utilizes the two visual features to generate better captions. GRIT replaces the CNN-based detector employed in previous methods with a DETR-based one, making it computationally faster. Moreover, its monolithic design consisting only of Transformers enables end-to-end training of the model. This innovative design and the integration of the dual visual features bring about significant performance improvement. The experimental results on several image captioning benchmarks show that GRIT outperforms previous methods in inference accuracy and speed.