Jul 31, 2016 · Abstract:Visual relationships capture a wide variety of interactions between pairs of objects in images (e.g. "man riding bicycle" and "man ...

Mar 16, 2023 · This work focuses on training a single visual relationship detector predicting over the union of label spaces from multiple datasets. Merging ...

We propose a model that uses this insight to train visual models for objects and predicates individually and later combines them together to predict multiple ...

This project deals with a visual relationship dataset that contains 5000 images, including 37,993 relations, 100 object categories, and 70 predicate categories.

Visual relationship detection (VRD) is one newly developed computer vision task aiming to recognize relations or interactions between objects in an image.

A curated list of visual relationship detection and related area (e.g. object detection, scene graph) resources, inspired by awesome-computer-vision. Contents.