Timezone: »

Generalized Focal Loss: Learning Qualified and Distributed Bounding Boxes for Dense Object Detection
Xiang Li · Wenhai Wang · Lijun Wu · Shuo Chen · Xiaolin Hu · Jun Li · Jinhui Tang · Jian Yang

Wed Dec 09 09:00 PM -- 11:00 PM (PST) @ Poster Session 4 #1152

One-stage detector basically formulates object detection as dense classification and localization (i.e., bounding box regression). The classification is usually optimized by Focal Loss and the box location is commonly learned under Dirac delta distribution. A recent trend for one-stage detectors is to introduce an \emph{individual} prediction branch to estimate the quality of localization, where the predicted quality facilitates the classification to improve detection performance. This paper delves into the \emph{representations} of the above three fundamental elements: quality estimation, classification and localization. Two problems are discovered in existing practices, including (1) the inconsistent usage of the quality estimation and classification between training and inference, and (2) the inflexible Dirac delta distribution for localization. To address the problems, we design new representations for these elements. Specifically, we merge the quality estimation into the class prediction vector to form a joint representation, and use a vector to represent arbitrary distribution of box locations. The improved representations eliminate the inconsistency risk and accurately depict the flexible distribution in real data, but contain \emph{continuous} labels, which is beyond the scope of Focal Loss. We then propose Generalized Focal Loss (GFL) that generalizes Focal Loss from its discrete form to the \emph{continuous} version for successful optimization. On COCO {\tt test-dev}, GFL achieves 45.0\% AP using ResNet-101 backbone, surpassing state-of-the-art SAPD (43.5\%) and ATSS (43.6\%) with higher or comparable inference speed.

Author Information

Xiang Li (NJUST)
Wenhai Wang (Nanjing University)
Lijun Wu (Sun Yat-sen University)
Shuo Chen (Nanjing University of Science and Technology)
Xiaolin Hu (Tsinghua University)
Jun Li (Nanjing University of Science and Technology)
Jinhui Tang (Nanjing University of Science and Technology)
Jian Yang (Nanjing University of Science and Technology)

More from the Same Authors