Multi-Stage Progressive Refinement and RoI Context Enhancement Network for Small Logo Detection

Songhui Zhao, Sujuan Hou

Abstract: Logo detection is a critical task in computer vision with a wide range of applications. In logo detection tasks, small logos occupy a limited number of pixels in the image due to their small size. Additionally, background clutter may have textures, colors, or shapes that are similar to small logos, further making it difficult for the detection algorithm to distinguish between the object and the background. To address this problem, we propose a Multi-stage Progressive Refinement and RoI (Region of Interest) Context Enhancement Network (MPRRCENet) for small logo detection. Specifically, a Multi-stage Progressive Refinement (MPR) module is proposed to progressive refine features with discriminative capability and promote the interaction of feature information. Furthermore, a RoI Context Enhancement (RCE) module is proposed that utilizes contextual information and channel modeling to enhance RoI features. Extensive experiments on four publicly available logo datasets demonstrate the effectiveness of our proposed method.