Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

A Generative Approach for Wikipedia-Scale Visual Entity Recognition

About

In this paper, we address web-scale visual entity recognition, specifically the task of mapping a given query image to one of the 6 million existing entities in Wikipedia. One way of approaching a problem of such scale is using dual-encoder models (eg CLIP), where all the entity names and query images are embedded into a unified space, paving the way for an approximate k-NN search. Alternatively, it is also possible to re-purpose a captioning model to directly generate the entity names for a given image. In contrast, we introduce a novel Generative Entity Recognition (GER) framework, which given an input image learns to auto-regressively decode a semantic and discriminative ``code'' identifying the target entity. Our experiments demonstrate the efficacy of this GER paradigm, showcasing state-of-the-art performance on the challenging OVEN benchmark. GER surpasses strong captioning, dual-encoder, visual matching and hierarchical classification baselines, affirming its advantage in tackling the complexities of web-scale recognition.

Mathilde Caron, Ahmet Iscen, Alireza Fathi, Cordelia Schmid• 2024

Related benchmarks

TaskDatasetResultRank
Fine-grained Entity RecognitionOVEN Entity 1.0 (test)
HM22.7
15
Visual Entity RecognitionOVEN
HM (Unseen)17.7
15
Visual Question AnsweringOVEN Query 1.0 (test)
HM6.3
15
Visual Entity RecognitionOVEN entity (test)
Top-1 Accuracy (Seen)29.1
11
Showing 4 of 4 rows

Other info

Follow for update