HOP+: History-Enhanced and Order-Aware Pre-Training for Vision-and-Language Navigation

Simple item page
dc.contributor.authorQiao, Y.
dc.contributor.authorQi, Y.
dc.contributor.authorHong, Y.
dc.contributor.authorYu, Z.
dc.contributor.authorWang, P.
dc.contributor.authorWu, Q.
dc.date.issued2023
dc.description.abstractRecent works attempt to employ pre-training in Vision-and-Language Navigation (VLN). However, these methods neglect the importance of historical contexts or ignore predicting future actions during pre-training, limiting the learning of visual-textual correspondence and the capability of decision-making. To address these problems, we present a history-enhanced and order-aware pre-training with the complementing fine-tuning paradigm (HOP+) for VLN. Specifically, besides the common Masked Language Modeling (MLM) and Trajectory-Instruction Matching (TIM) tasks, we design three novel VLN-specific proxy tasks: Action Prediction with History (APH) task, Trajectory Order Modeling (TOM) task and Group Order Modeling (GOM) task. APH task takes into account the visual perception trajectory to enhance the learning of historical knowledge as well as action prediction. The two temporal visualtextual alignment tasks, TOM and GOM further improve the agent’s ability to order reasoning. Moreover, we design a memory network to address the representation inconsistency of history context between the pre-training and the fine-tuning stages. The memory network effectively selects and summarizes historical information for action prediction during fine-tuning, without costing huge extra computation consumption for downstream VLN tasks. HOP+ achieves new state-of-the-art performance on four downstream VLN tasks (R2R, REVERIE, RxR, and NDH), which demonstrates the effectiveness of our proposed method.
dc.description.statementofresponsibilityYanyuan Qiao, Yuankai Qi, Yicong Hong, Zheng Yu, Peng Wang, and Qi Wu
dc.identifier.citationIEEE Transactions on Pattern Analysis and Machine Intelligence, 2023; 45(7):8524-8537
dc.identifier.doi10.1109/tpami.2023.3234243
dc.identifier.issn0162-8828
dc.identifier.issn2160-9292
dc.identifier.orcidQiao, Y. [0000-0002-5606-0702]
dc.identifier.orcidWu, Q. [0000-0003-3631-256X]
dc.identifier.urihttps://hdl.handle.net/2440/139132
dc.language.isoen
dc.publisherInstitute of Electrical and Electronics Engineers (IEEE)
dc.relation.granthttp://purl.org/au-research/grants/arc/DE190100539
dc.rights© 2023 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
dc.source.urihttps://doi.org/10.1109/tpami.2023.3234243
dc.subjectVision-and-language navigation; pre-training; memory networks
dc.titleHOP+: History-Enhanced and Order-Aware Pre-Training for Vision-and-Language Navigation
dc.typeJournal article
pubs.publication-statusPublished

Files

Collections

Australian Institute for Machine Learning publications