Transferable Representation Learning in Vision-and-Language Navigation
This work addresses the challenge of building more effective navigation agents that integrate vision and language for tasks like R2R, representing an incremental improvement over existing competitive methods.
The paper tackles improving agent performance in Vision-and-Language Navigation (VLN) by adapting pre-trained vision and language representations to in-domain tasks, specifically cross-modal sequence alignment and coherence, resulting in enhanced success rates as measured by the SPL metric on the Room-to-Room (R2R) benchmark.
Vision-and-Language Navigation (VLN) tasks such as Room-to-Room (R2R) require machine agents to interpret natural language instructions and learn to act in visually realistic environments to achieve navigation goals. The overall task requires competence in several perception problems: successful agents combine spatio-temporal, vision and language understanding to produce appropriate action sequences. Our approach adapts pre-trained vision and language representations to relevant in-domain tasks making them more effective for VLN. Specifically, the representations are adapted to solve both a cross-modal sequence alignment and sequence coherence task. In the sequence alignment task, the model determines whether an instruction corresponds to a sequence of visual frames. In the sequence coherence task, the model determines whether the perceptual sequences are predictive sequentially in the instruction-conditioned latent space. By transferring the domain-adapted representations, we improve competitive agents in R2R as measured by the success rate weighted by path length (SPL) metric.