TY - JOUR AU - Liu, Yudong AU - Wang, Yongtao AU - Wang, Siwei AU - Liang, Tingting AU - Zhao, Qijie AU - Tang, Zhi AU - Ling, Haibin PY - 2020/04/03 Y2 - 2024/03/28 TI - CBNet: A Novel Composite Backbone Network Architecture for Object Detection JF - Proceedings of the AAAI Conference on Artificial Intelligence JA - AAAI VL - 34 IS - 07 SE - AAAI Technical Track: Vision DO - 10.1609/aaai.v34i07.6834 UR - https://ojs.aaai.org/index.php/AAAI/article/view/6834 SP - 11653-11660 AB - <p>In existing CNN based detectors, the backbone network is a very important component for basic feature<sup>1</sup> extraction, and the performance of the detectors highly depends on it. In this paper, we aim to achieve better detection performance by building a more powerful backbone from existing ones like ResNet and ResNeXt. Specifically, we propose a novel strategy for assembling multiple identical backbones by composite connections between the adjacent backbones, to form a more powerful backbone named <em>Composite Backbone Network</em> (CBNet). In this way, CBNet iteratively feeds the output features of the previous backbone, namely <em>high-level features</em>, as part of input features to the succeeding backbone, in a stage-by-stage fashion, and finally the feature maps of the last backbone (named <em>Lead Backbone</em>) are used for object detection. We show that CBNet can be very easily integrated into most state-of-the-art detectors and significantly improve their performances. For example, it boosts the mAP of FPN, Mask R-CNN and Cascade R-CNN on the COCO dataset by about 1.5 to 3.0 points. Moreover, experimental results show that the instance segmentation results can be improved as well. Specifically, by simply integrating the proposed CBNet into the baseline detector Cascade Mask R-CNN, we achieve a new state-of-the-art result on COCO dataset (mAP of 53.3) with a single model, which demonstrates great effectiveness of the proposed CBNet architecture. Code will be made available at https://github.com/PKUbahuangliuhe/CBNet.</p> ER -