English

VISITRON: Visual Semantics-Aligned Interactively Trained Object-Navigator

Computer Vision and Pattern Recognition 2022-03-17 v2 Artificial Intelligence Computation and Language Machine Learning Robotics

Abstract

Interactive robots navigating photo-realistic environments need to be trained to effectively leverage and handle the dynamic nature of dialogue in addition to the challenges underlying vision-and-language navigation (VLN). In this paper, we present VISITRON, a multi-modal Transformer-based navigator better suited to the interactive regime inherent to Cooperative Vision-and-Dialog Navigation (CVDN). VISITRON is trained to: i) identify and associate object-level concepts and semantics between the environment and dialogue history, ii) identify when to interact vs. navigate via imitation learning of a binary classification head. We perform extensive pre-training and fine-tuning ablations with VISITRON to gain empirical insights and improve performance on CVDN. VISITRON's ability to identify when to interact leads to a natural generalization of the game-play mode introduced by Roman et al. (arXiv:2005.00728) for enabling the use of such models in different environments. VISITRON is competitive with models on the static CVDN leaderboard and attains state-of-the-art performance on the Success weighted by Path Length (SPL) metric.

Keywords

Cite

@article{arxiv.2105.11589,
  title  = {VISITRON: Visual Semantics-Aligned Interactively Trained Object-Navigator},
  author = {Ayush Shrivastava and Karthik Gopalakrishnan and Yang Liu and Robinson Piramuthu and Gokhan Tür and Devi Parikh and Dilek Hakkani-Tür},
  journal= {arXiv preprint arXiv:2105.11589},
  year   = {2022}
}

Comments

Accepted at Findings of the Annual Meeting of the Association for Computational Linguistics (ACL) 2022, previous version accepted at Visually Grounded Interaction and Language (ViGIL) Workshop at NAACL 2021

R2 v1 2026-06-24T02:25:36.972Z