3A-YOLO: New Real-Time Object Detectors with Triple Discriminative Awareness and Coordinated Representations

TL;DR

This paper introduces 3A-YOLO, a real-time object detector that hierarchically integrates multiple attention mechanisms to enhance discriminative awareness and intermediate feature representations, leading to improved detection performance.

Contribution

It proposes a novel TDA-YOLO Module that unifies scale, spatial, and task awareness, and introduces coordinated learning of intermediate features for better object detection.

Findings

Outperforms existing YOLO variants on COCO and VOC benchmarks.

Achieves higher accuracy with real-time inference speed.

Demonstrates the effectiveness of hierarchical attention mechanisms.

Abstract

Recent research on real-time object detectors (e.g., YOLO series) has demonstrated the effectiveness of attention mechanisms for elevating model performance. Nevertheless, existing methods neglect to unifiedly deploy hierarchical attention mechanisms to construct a more discriminative YOLO head which is enriched with more useful intermediate features. To tackle this gap, this work aims to leverage multiple attention mechanisms to hierarchically enhance the triple discriminative awareness of the YOLO detection head and complementarily learn the coordinated intermediate representations, resulting in a new series detectors denoted 3A-YOLO. Specifically, we first propose a new head denoted TDA-YOLO Module, which unifiedly enhance the representations learning of scale-awareness, spatial-awareness, and task-awareness. Secondly, we steer the intermediate features to coordinately learn the inter-channel relationships and precise positional information. Finally, we perform neck network improvements followed by introducing various tricks to boost the adaptability of 3A-YOLO. Extensive experiments across COCO and VOC benchmarks indicate the effectiveness of our detectors.