计算模型的参数量(Params)和运算量(FLOPs)

Compute the parameters and FLOPs of the model using thop.

Objects as Points

CenterNet：把目标检测视为中心点检测.

Argmax与SoftArgmax

Implement SoftArgmax with Pytorch.

Focal Loss for Dense Object Detection

RetinaNet：密集目标检测的焦点损失.

SSD: Single Shot MultiBox Detector

SSD：单次多边界框检测器.

YOLOv3: An Incremental Improvement

YOLOv3：对YOLOv2模型的增量式改进.