Composing Text and Image for Image Retrieval - An Empirical Odyssey
This addresses the need for more flexible image retrieval systems that can handle multimodal queries, though it is incremental as it builds on existing embedding and composition techniques.
The paper tackles the problem of image retrieval using a query composed of an image and text describing modifications, proposing a method to learn a similarity metric between target images and composed source image-text features. It demonstrates improved performance over existing approaches on three datasets, including Fashion-200k, MIT-States, and a new synthetic dataset based on CLEVR.
In this paper, we study the task of image retrieval, where the input query is specified in the form of an image plus some text that describes desired modifications to the input image. For example, we may present an image of the Eiffel tower, and ask the system to find images which are visually similar but are modified in small ways, such as being taken at nighttime instead of during the day. To tackle this task, we learn a similarity metric between a target image and a source image plus source text, an embedding and composing function such that target image feature is close to the source image plus text composition feature. We propose a new way to combine image and text using such function that is designed for the retrieval task. We show this outperforms existing approaches on 3 different datasets, namely Fashion-200k, MIT-States and a new synthetic dataset we create based on CLEVR. We also show that our approach can be used to classify input queries, in addition to image retrieval.