Generative retrieval is an emerging approach in information retrieval that generates identifiers (IDs) of target data based on a query, providing an efficient alternative to traditional embedding-based retrieval methods. However, existing models are task-specific and fall short of embedding-based retrieval in performance. This paper proposes GENIUS, a universal generative retrieval framework supporting diverse tasks across multiple modalities and domains. At its core, GENIUS introduces modality-decoupled semantic quantization, transforming multimodal data into discrete IDs encoding both modality and semantics. Moreover, to enhance generalization, we propose a query augmentation that interpolates between a query and its target, allowing GENIUS to adapt to varied query forms. Evaluated on the M-BEIR benchmark, it surpasses prior generative methods by a clear margin. Unlike embedding-based retrieval, GENIUS consistently maintains high retrieval speed across database size, with competitive performance across multiple benchmarks. With additional re-ranking, GENIUS often achieves results close to those of embedding-based methods while preserving efficiency.

Image to image matching has been well studied in the computer vision community. Previous studies mainly focus on training a deep metric learning model matching visual patterns between the query image and gallery images. In this study, we show that pure image-to- image matching suffers from false positives caused by matching to local visual patterns. To alleviate this issue, we propose to leverage recent advances in vision-language pretraining research. Specifically, we introduce additional image-text alignment losses into deep metric learning, which serve as constraints to the image-to-image matching loss. With additional alignments between the text (e.g., product title) and image pairs, the model can learn concepts from both modalities explicitly, which avoids matching low-level visual features. We progressively develop two variants, a 3-tower and a 4-tower model, where the latter takes one more short text query input. Through extensive experiments, we show that this change leads to a substantial improvement to the image to image matching problem. We further leveraged this model for multimodal search, which takes both image and reformulation text queries to improve search quality. Both offline and online experiments show strong improvements on the main metrics. Specifically, we see 4.95% relative improvement on image matching click through rate with the 3-tower model and 1.13% further improvement from the 4-tower model.

Multimodal data is available in many applications like e-commerce production listings, social media posts and short videos. However, existing algorithms dealing with those types of data still focus on uni-modal representation learning by vision-language alignment and cross-modal retrieval. In this workshop, we target to bring a new retrieval problem where both queries and documents are multimodal. With the popularity of vision language modeling, large language models (LLMs), retrieval augmented generation (RAG), and multimodal LLM, we see a lot of new opportunities for multimodal representation and retrieval tasks. This event will be a comprehensive half-day workshop focusing on the subject of multimodal representation and retrieval. The agenda includes keynote speeches, oral presentations, and an interactive panel discussion.

In this paper, we introduce Shop the Look, a web-scale fashion and home product visual search system deployed at Amazon. Building such a system poses great challenges to both science and engineering practices. We leverage large-scale image data from the Amazon product catalog and adopt effective strategies to reduce the human effort required to annotate data. By employing state-of-the-art computer vision techniques, we train detection, recognition, and feature extraction models to bridge the domain gap between in-the-wild query images and product images which are taken under controlled settings. Our system is designed to achieve a balance between result accuracy and efficiency. The run-time service is optimized to provide retrieval results to users with low-latency. The scalable offline index-building pipeline adapts to the dynamic Amazon catalog that contains billions of products. We present both quantitative and qualitative evaluation results to demonstrate the performance of our system. We believe that the fast-growing Shop the Look service is shaping the way that customers shop on Amazon.

Proxy-based metric learning losses are superior to pair-based losses due to their fast convergence and low training complexity. However, existing proxy-based losses focus on learning class-discriminative features while overlooking the commonalities shared across classes which are potentially useful in describing and matching samples. Moreover, they ignore the implicit hierarchy of categories in real-world datasets, where similar subordinate classes can be grouped together. In this paper, we present a framework that leverages this implicit hierarchy by imposing a hierarchical structure on the proxies and can be used with any existing proxy-based loss. This allows our model to capture both class-discriminative features and class-shared characteristics without breaking the implicit data hierarchy. We evaluate our method on five established image retrieval datasets such as In-Shop and SOP. Results demonstrate that our hierarchical proxy-based loss framework improves the performance of existing proxy-based losses, especially on large datasets which exhibit strong hierarchical structure.

We present a system for making community driven websites easily accessible from the latest mobile devices. Many of these new devices contain an ensemble of sensors such as cameras, GPS and inertial sensors. We demonstrate how these new sensors can be used to bring the information contained in sites like Wikipedia to users in a much more immersive manner than text or maps. We have collected a large database of images and articles from Wikipedia and show how a user can query this database by simply snapping a photo. Our system uses the location sensors to assist with image matching and the inertial sensors to provide a unique and intuitive user interface for browsing results.

This paper describes a learning-based method for counting people in crowds from a single camera. Our method takes into account feature normalization to deal with perspective projection and different camera orientation. Thus, our system is trained to be viewpoint invariant and can be deployed with minimal setup for a new site. This is achieved by applying background subtraction and edge detection to each frame and extracting edge orientation and blob size histograms as features. A homography is computed between the ground plane and the image plane coordinates for the region of interest (ROI). A density map that measures the relative size of individuals and a global scale measuring camera orientation are also estimated and used for feature normalization. The relationship between the feature histograms and the number of pedestrians in the crowds is learned from labeled training data. The two training methods used in the current system are linear fitting and neural networks. Experimental results from different sites with different camera orientation demonstrate the performance and the potential of our method. 1