Image Captioning via Dynamic Path Customization

TL;DR

This paper introduces a dynamic network architecture for image captioning that customizes its structure on a per-input basis, improving caption quality by leveraging spatial and channel information for more discriminative and accurate descriptions.

Contribution

The paper proposes the Dynamic Transformer Network (DTNet) with a novel Spatial-Channel Joint Router for adaptive path customization, advancing beyond static, handcrafted models.

Findings

Achieves state-of-the-art results on MS-COCO dataset.

Demonstrates improved caption quality through dynamic routing.

Validates effectiveness across multiple evaluation splits.

Abstract

This paper explores a novel dynamic network for vision and language tasks, where the inferring structure is customized on the fly for different inputs. Most previous state-of-the-art approaches are static and hand-crafted networks, which not only heavily rely on expert knowledge, but also ignore the semantic diversity of input samples, therefore resulting in suboptimal performance. To address these issues, we propose a novel Dynamic Transformer Network (DTNet) for image captioning, which dynamically assigns customized paths to different samples, leading to discriminative yet accurate captions. Specifically, to build a rich routing space and improve routing efficiency, we introduce five types of basic cells and group them into two separate routing spaces according to their operating domains, i.e., spatial and channel. Then, we design a Spatial-Channel Joint Router (SCJR), which endows the model with the capability of path customization based on both spatial and channel information of the input sample. To validate the effectiveness of our proposed DTNet, we conduct extensive experiments on the MS-COCO dataset and achieve new state-of-the-art performance on both the Karpathy split and the online test server.