The paper "A Fair Ranking and New Model for Panoptic Scene Graph Generation" by Julian Lorenz, Alexander Pest, Daniel Kienzle, Katja Ludwig, and Rainer Lienhart has been accepted for ECCV 2024.

The authors discuss significant flaws in commonly used evaluation protocols for Panoptic Scene Graph Generation. They present a solution to this problem and evaluate existing publications based on the new findings.
Finally, a new state-of-the-art architecture for Panoptic Scene Graph Generation is presented.

More information can be found here: https://lorjul.github.io/fair-psgg/

In panoptic scene graph generation (PSGG), models retrieve interactions between objects in an image which are grounded by panoptic segmentation masks.  
Previous evaluations on panoptic scene graphs have been subject to an erroneous evaluation protocol where multiple masks for the same object can lead to multiple relation distributions per mask-mask pair. This can be exploited to increase the final score. We correct this flaw and provide a fair ranking over a wide range of existing PSGG models.  
The observed scores for existing methods increase by up to 7.4 mR@50 for all two-stage methods, while dropping by up to 19.3 mR@50 for all one-stage methods, highlighting the importance of a correct evaluation. Contrary to recent publications, we show that existing two-stage methods are competitive to one-stage methods. Building on this, we introduce the Decoupled SceneFormer (DSFormer), a novel two-stage model that outperforms all existing scene graph models by a large margin of +11 mR@50 and +10 mNgR@50 on the corrected evaluation, thus setting a new SOTA. As a core design principle, DSFormer encodes subject and object masks directly into feature space.