REACT++: Efficient Cross-Attention for Real-Time Scene Graph Generation

TL;DR

REACT++ is a new model for real-time scene graph generation that balances speed and accuracy by using efficient feature extraction and cross-attention, outperforming previous models in inference speed and relation prediction.

Contribution

It introduces REACT++, a novel approach that improves inference speed and relation prediction accuracy while maintaining object detection performance in real-time SGG.

Findings

REACT++ is 20% faster than previous models.

REACT++ achieves a 10% average gain in relation prediction accuracy.

REACT++ maintains high object detection performance.

Abstract

Scene Graph Generation (SGG) is a task that encodes visual relationships between objects in images as graph structures. SGG shows significant promise as a foundational component for downstream tasks, such as reasoning for embodied agents. To enable real-time applications, SGG must address the trade-off between performance and inference speed. However, current methods tend to focus on one of the following: (1) improving relation prediction accuracy, (2) enhancing object detection accuracy, or (3) reducing latency, without aiming to balance all three objectives simultaneously. To address this limitation, we build on the powerful Real-time Efficiency and Accuracy Compromise for Tradeoffs in Scene Graph Generation (REACT) architecture and propose REACT++, a new state-of-the-art model for real-time SGG. By leveraging efficient feature extraction and subject-to-object cross-attention within the prototype space, REACT++ balances latency and representational power. REACT++ achieves the highest inference speed among existing SGG models, improving relation prediction accuracy without sacrificing object detection performance. Compared to the previous REACT version, REACT++ is 20% faster with a gain of 10% in relation prediction accuracy on average. The code is available at https://github.com/Maelic/SGG-Benchmark.