EnvEdit: Environment Editing for Vision-and-Language Navigation

TL;DR

EnvEdit is a data augmentation technique for Vision-and-Language Navigation that creates diverse environments by editing existing ones, improving agent generalization and achieving state-of-the-art results.

Contribution

The paper introduces EnvEdit, a novel environment editing method for data augmentation in VLN, enhancing generalization to unseen environments.

Findings

Significant performance improvements on Room-to-Room and Room-Across-Room datasets.

Achieves new state-of-the-art on VLN test leaderboard.

Ensemble of edited environments further boosts performance.

Abstract

In Vision-and-Language Navigation (VLN), an agent needs to navigate through the environment based on natural language instructions. Due to limited available data for agent training and finite diversity in navigation environments, it is challenging for the agent to generalize to new, unseen environments. To address this problem, we propose EnvEdit, a data augmentation method that creates new environments by editing existing environments, which are used to train a more generalizable agent. Our augmented environments can differ from the seen environments in three diverse aspects: style, object appearance, and object classes. Training on these edit-augmented environments prevents the agent from overfitting to existing environments and helps generalize better to new, unseen environments. Empirically, on both the Room-to-Room and the multi-lingual Room-Across-Room datasets, we show that our proposed EnvEdit method gets significant improvements in all metrics on both pre-trained and non-pre-trained VLN agents, and achieves the new state-of-the-art on the test leaderboard. We further ensemble the VLN agents augmented on different edited environments and show that these edit methods are complementary. Code and data are available at https://github.com/jialuli-luka/EnvEdit