Graph-Based Captioning: Enhancing Visual Descriptions by Interconnecting Region Captions

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Hsieh, Yu-Guan, Hsieh, Cheng-Yu, Yeh, Shih-Ying, Béthune, Louis, Ansari, Hadi Pour, Vasu, Pavan Kumar Anasosalu, Li, Chun-Liang, Krishna, Ranjay, Tuzel, Oncel, Cuturi, Marco
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866909513173434368
author Hsieh, Yu-Guan
Hsieh, Cheng-Yu
Yeh, Shih-Ying
Béthune, Louis
Ansari, Hadi Pour
Vasu, Pavan Kumar Anasosalu
Li, Chun-Liang
Krishna, Ranjay
Tuzel, Oncel
Cuturi, Marco
author_facet Hsieh, Yu-Guan
Hsieh, Cheng-Yu
Yeh, Shih-Ying
Béthune, Louis
Ansari, Hadi Pour
Vasu, Pavan Kumar Anasosalu
Li, Chun-Liang
Krishna, Ranjay
Tuzel, Oncel
Cuturi, Marco
contents Humans describe complex scenes with compositionality, using simple text descriptions enriched with links and relationships. While vision-language research has aimed to develop models with compositional understanding capabilities, this is not reflected yet in existing datasets which, for the most part, still use plain text to describe images. In this work, we propose a new annotation strategy, graph-based captioning (GBC) that describes an image using a labeled graph structure, with nodes of various types. The nodes in GBC are created through a two-stage process: first, identifying and describing entity nodes; second, linking these nodes by highlighting \textit{compositions} and \textit{relations} among them. Since \textit{all} GBC nodes hold plain text descriptions, GBC retains the flexibility found in natural language, but can also encode hierarchical information in its edges. We demonstrate that GBC can be produced automatically, using off-the-shelf multimodal LLMs and object detection models, by building a new dataset GBC10M that gathers GBC annotations for about 10M images of the CC12M dataset. Through CLIP training on GBC10M, we show that leveraging GBC nodes' annotations -- particularly those in composition and relation nodes -- significantly boosts the model's performance across various benchmarks compared to when other annotations are used. To further explore the opportunities provided by GBC, we also investigate the use of GBC as middleware for text-to-image generation, and show the extra benefits of incorporating the graph structure in this task. Our code and datasets are released at https://github.com/apple/ml-gbc and https://huggingface.co/graph-based-captions.
format Preprint
id arxiv_https___arxiv_org_abs_2407_06723
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Graph-Based Captioning: Enhancing Visual Descriptions by Interconnecting Region Captions
Hsieh, Yu-Guan
Hsieh, Cheng-Yu
Yeh, Shih-Ying
Béthune, Louis
Ansari, Hadi Pour
Vasu, Pavan Kumar Anasosalu
Li, Chun-Liang
Krishna, Ranjay
Tuzel, Oncel
Cuturi, Marco
Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
Humans describe complex scenes with compositionality, using simple text descriptions enriched with links and relationships. While vision-language research has aimed to develop models with compositional understanding capabilities, this is not reflected yet in existing datasets which, for the most part, still use plain text to describe images. In this work, we propose a new annotation strategy, graph-based captioning (GBC) that describes an image using a labeled graph structure, with nodes of various types. The nodes in GBC are created through a two-stage process: first, identifying and describing entity nodes; second, linking these nodes by highlighting \textit{compositions} and \textit{relations} among them. Since \textit{all} GBC nodes hold plain text descriptions, GBC retains the flexibility found in natural language, but can also encode hierarchical information in its edges. We demonstrate that GBC can be produced automatically, using off-the-shelf multimodal LLMs and object detection models, by building a new dataset GBC10M that gathers GBC annotations for about 10M images of the CC12M dataset. Through CLIP training on GBC10M, we show that leveraging GBC nodes' annotations -- particularly those in composition and relation nodes -- significantly boosts the model's performance across various benchmarks compared to when other annotations are used. To further explore the opportunities provided by GBC, we also investigate the use of GBC as middleware for text-to-image generation, and show the extra benefits of incorporating the graph structure in this task. Our code and datasets are released at https://github.com/apple/ml-gbc and https://huggingface.co/graph-based-captions.
title Graph-Based Captioning: Enhancing Visual Descriptions by Interconnecting Region Captions
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2407.06723