Discriminative Semantic Feature Pyramid Network with Guided Anchoring for Logo Detection

Logo detection is a technology that identifies logos in images and returns their locations. With logo technology, brands can check how often are displayed on social media platforms elsewhere online they appear. It has received lot of attention for its wide applications across different sectors, such as brand identity protection, product management, duration monitoring. Particularly, offer various benefits companies to help measure coverage, track perception, secure value, increase the effectiveness marketing campaigns build awareness more effectively. However, compared with general object detection, challenging due existence both small objects large aspect ratio objects. In this paper, we propose novel approach, named Discriminative Semantic Feature Pyramid Network Guided Anchoring (DSFP-GA), which address these challenges via aggregating semantic information generating anchor boxes. More specifically, our approach mainly consists two components, namely (DSFP) (GA). The former proposed fuse features into low-level feature maps obtain discriminative representation objects, while latter further integrated DSFP generate boxes detecting Extensive experimental results four benchmarks demonstrate DSFP-GA. Moreover, conduct visual analysis ablation studies illustrate strength DSFP-GA when