Embodied Image Captioning: Self-supervised Learning Agents for Spatially Coherent Image Descriptions

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Galliena, Tommaso, Apicella, Tommaso, Rosa, Stefano, Morerio, Pietro, Del Bue, Alessio, Natale, Lorenzo
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916952549621760
author Galliena, Tommaso
Apicella, Tommaso
Rosa, Stefano
Morerio, Pietro
Del Bue, Alessio
Natale, Lorenzo
author_facet Galliena, Tommaso
Apicella, Tommaso
Rosa, Stefano
Morerio, Pietro
Del Bue, Alessio
Natale, Lorenzo
contents We present a self-supervised method to improve an agent's abilities in describing arbitrary objects while actively exploring a generic environment. This is a challenging problem, as current models struggle to obtain coherent image captions due to different camera viewpoints and clutter. We propose a three-phase framework to fine-tune existing captioning models that enhances caption accuracy and consistency across views via a consensus mechanism. First, an agent explores the environment, collecting noisy image-caption pairs. Then, a consistent pseudo-caption for each object instance is distilled via consensus using a large language model. Finally, these pseudo-captions are used to fine-tune an off-the-shelf captioning model, with the addition of contrastive learning. We analyse the performance of the combination of captioning models, exploration policies, pseudo-labeling methods, and fine-tuning strategies, on our manually labeled test set. Results show that a policy can be trained to mine samples with higher disagreement compared to classical baselines. Our pseudo-captioning method, in combination with all policies, has a higher semantic similarity compared to other existing methods, and fine-tuning improves caption accuracy and consistency by a significant margin. Code and test set annotations available at https://hsp-iit.github.io/embodied-captioning/
format Preprint
id arxiv_https___arxiv_org_abs_2504_08531
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Embodied Image Captioning: Self-supervised Learning Agents for Spatially Coherent Image Descriptions
Galliena, Tommaso
Apicella, Tommaso
Rosa, Stefano
Morerio, Pietro
Del Bue, Alessio
Natale, Lorenzo
Computer Vision and Pattern Recognition
Robotics
We present a self-supervised method to improve an agent's abilities in describing arbitrary objects while actively exploring a generic environment. This is a challenging problem, as current models struggle to obtain coherent image captions due to different camera viewpoints and clutter. We propose a three-phase framework to fine-tune existing captioning models that enhances caption accuracy and consistency across views via a consensus mechanism. First, an agent explores the environment, collecting noisy image-caption pairs. Then, a consistent pseudo-caption for each object instance is distilled via consensus using a large language model. Finally, these pseudo-captions are used to fine-tune an off-the-shelf captioning model, with the addition of contrastive learning. We analyse the performance of the combination of captioning models, exploration policies, pseudo-labeling methods, and fine-tuning strategies, on our manually labeled test set. Results show that a policy can be trained to mine samples with higher disagreement compared to classical baselines. Our pseudo-captioning method, in combination with all policies, has a higher semantic similarity compared to other existing methods, and fine-tuning improves caption accuracy and consistency by a significant margin. Code and test set annotations available at https://hsp-iit.github.io/embodied-captioning/
title Embodied Image Captioning: Self-supervised Learning Agents for Spatially Coherent Image Descriptions
topic Computer Vision and Pattern Recognition
Robotics
url https://arxiv.org/abs/2504.08531