This paper presents the Global Feature Pyramid Network (GFPNet), an improved version of PAFPN that uses global data to enhance target detection. The GFPNet addresses the issue of increased false detections and missed targets by incorporating global information interaction, which has been overlooked in previous methodologies. This is done using a lightweight MLP to capture global feature information, the VNC encoder to process these features, and a parallel learnable mechanism to extract intra-layer features from the input image. The PAFPN method is retained to enable inter-layer feature interaction, extracting rich feature details across various levels. The GFPNet not only focuses on inter-layer feature information but also captures global feature details, fostering intra-layer feature interaction and generating a more comprehensive and impactful feature representation.
Publication date: 19 Dec 2023
Project Page: www.elsevier.com
Paper: https://arxiv.org/pdf/2312.11231
