Arnab Dhua

Jun 13, 2024 | amazon.science | Shasha Li |Ming Du |Arnab Dhua |Shuai Tang

Vision-language transformer models play a pivotal role in e-commerce product search. When using product description (e.g. product title) and product image pairs to train such models, there are often non-visual-descriptive text attributes in the product description, which makes the visual textual alignment challenging. We introduce MultiModal Learning with online Token Pruning (MML-TP).

Apr 16, 2024 | amazon.science | Michael Huang |Xinliang Zhu |Arnab Dhua |Oriol Barbany Mayor

Multimodal search has become increasingly important in providing users with a natural and effective way to ex-press their search intentions. Images offer fine-grained details of the desired products, while text allows for easily incorporating search modifications. However, some existing multimodal search systems are unreliable and fail to address simple queries.