SCoRD: Subject-Conditional Relation Detection with Text-Augmented Data

TL;DR

This paper introduces SCoRD, a subject-conditional relation detection method that predicts relations and object locations conditioned on a subject, demonstrating improved performance and generalization on a new challenging benchmark based on the Open Images dataset.

Contribution

The paper presents a novel auto-regressive model for subject-conditional relation detection and introduces the OIv6-SCoRD benchmark with distribution shift, improving upon previous scene graph prediction methods.

Findings

Achieved recall@3 of 83.8% for relation-object predictions, outperforming previous methods.

Leveraged textual captions to improve relation-object and object-box prediction generalization.

Demonstrated better handling of distribution shifts in the new benchmark.

Abstract

We propose Subject-Conditional Relation Detection SCoRD, where conditioned on an input subject, the goal is to predict all its relations to other objects in a scene along with their locations. Based on the Open Images dataset, we propose a challenging OIv6-SCoRD benchmark such that the training and testing splits have a distribution shift in terms of the occurrence statistics of $\langle$subject, relation, object$\rangle$ triplets. To solve this problem, we propose an auto-regressive model that given a subject, it predicts its relations, objects, and object locations by casting this output as a sequence of tokens. First, we show that previous scene-graph prediction methods fail to produce as exhaustive an enumeration of relation-object pairs when conditioned on a subject on this benchmark. Particularly, we obtain a recall@3 of 83.8% for our relation-object predictions compared to the 49.75% obtained by a recent scene graph detector. Then, we show improved generalization on both relation-object and object-box predictions by leveraging during training relation-object pairs obtained automatically from textual captions and for which no object-box annotations are available. Particularly, for $\langle$subject, relation, object$\rangle$ triplets for which no object locations are available during training, we are able to obtain a recall@3 of 33.80% for relation-object pairs and 26.75% for their box locations.