Gazelle: A Multimodal Learning System Robust to Missing Modalities

Shah Nawaz, Muhammad Saad Saeed, Muhammad Zaigham Zaheer, Muhammad Haris Khan, Karthik Nandakumar, Muhammad Haroon Yousaf, Hassan Sajjad, Tom De Schepper, Markus Schedl

Keywords: multimodal, multimodal classification, missing modality

Abstract: Typical multimodal classification systems exhibit deteriorated performance if one or more modalities are missing at test time. In this work, we propose a robust multimodal classification system, Gazelle, which is less susceptible to missing modalities. It consists of a single-branch network sharing weights across multiple modalities to learn intermodal representations. It introduces a novel training scheme featuring a modality switch mechanism over input embeddings extracted using modality-specific networks to maximise performance as well as robustness to missing modalities. Extensive experiments are performed on four challenging datasets including textual-visual (UPMC Food-$101$, Hateful Memes, Ferramenta) and audio-visual modalities (VoxCeleb$1$). Gazelle achieved superior performance when all modalities are present as well as in the case of missing modalities compared to the existing state-of-the-art methods.

Primary Area: general machine learning (i.e., none of the above)

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 3373