Category-Oriented Representation Learning for Image to Multi-Modal Retrieval
About
The rise of multi-modal search requests from users has highlighted the importance of multi-modal retrieval (i.e. image-to-text or text-to-image retrieval), yet the more complex task of image-to-multi-modal retrieval, crucial for many industry applications, remains under-explored. To address this gap and promote further research, we introduce and define the concept of Image-to-Multi-Modal Retrieval (IMMR), a process designed to retrieve rich multi-modal (i.e. image and text) documents based on image queries. We focus on representation learning for IMMR and analyze three key challenges for it: 1) skewed data and noisy label in real-world industrial data, 2) the information-inequality between image and text modality of documents when learning representations, 3) effective and efficient training in large-scale industrial contexts. To tackle the above challenges, we propose a novel framework named organizing categories and learning by classification for retrieval (OCLEAR). It consists of three components: 1) a novel category-oriented data governance scheme coupled with a large-scale classification-based learning paradigm, which handles the skewed and noisy data from a data perspective. 2) model architecture specially designed for multi-modal learning, where information-inequality between image and text modality of documents is considered for modality fusion. 3) a hybrid parallel training approach for tackling large-scale training in industrial scenario. The proposed framework achieves SOTA performance on public datasets and has been deployed in a real-world industrial e-commence system, leading to significant business growth. Code will be made publicly available.