近几年目标检测综述(更新中)
(通过公众号,交流群收集一些目标检测网络)
+++++++++++++++++++++++++++++++++++++++++++++++++
- CBNet -> COCO : 53.3AP
- 名称:《CBNet: A Novel Composite Backbone Network Architecture for Object Detection》(AAAI 2020)
- 论文:(https://arxiv.org/abs/1909.03625)
- 亮点:COCO数据集霸榜模型,新颖的骨干网络(复合式骨干网)
- DetectoRS -> COCO: 54.7AP
- 名称:《DetectoRS: Detecting Objects with Recursive Feature Pyramid and Switchable Atrous Convolution》(3 Jun 2020 arXiv)
- 论文:(https://arxiv.org/abs/2006.02334)
- 亮点:递归特征金字塔网络(Recursive Feature Pyramid ,RFP),可切换空洞卷积(Switchable Atrous Convolution, SAC)
- 显著性目标检测:从图像中获得最显著的物体
- 论文/代码链接:https://github.com/jiwei0921/SOD-CNNs-based-code-summary-
- 可用于抠图软件、机器人避障、自动驾驶、医学诊断等。
- BorderDet: Border Feature for Dense Object Detection
- 论文链接:https://arxiv.org/pdf/2007.11056v1.pdf
- 论文代码:https://github.com/Megvii-BaseDetection/BorderDet
- 关键词:物体检测、边界特征
- 针对问题:传统的预定框检测物体的方法,它们往往可能缺乏有用的边界信息,从而不利于进行精准地定位。
- YOLOF:速度更快,精度堪比YOLOv4/RetinaNet,开源