This function is taken from the original tf repo. It ensures that all layers have a channel number that is divisible by 8 It can be seen

1x1 convolution

3x3 convolution with padding

任意四点poly nms.取出nms后的边框的索引 @param dets: shape(detection_num, [poly]) 原始图像中的检测出的目标数量 @param scores: shape(detection_num, 1) @param thre

calculate random scaling factor Args: s (float): range of the random scale. Returns: random scaling factor (float) whose

`Resnet50` is a function that returns a `Backbone` object with the following parameters: - `name`: "resnet50" - `pretrained`: False - `di

计算每个类别的ap值 :param iou_thresh: :param use_07_metric: :return:dict{cls:ap}

预测框进行过滤，去除尺度不合理的框

Note ：nn.Conv2d nn.BatchNorm2d'initing modes are uniform