Loads class labels at 'path'

Parses and loads the weights stored in 'weights_path

Returns the IoU of two bounding boxes

Removes detections with lower object confidence score than 'conf_thres' and performs Non-Maximum Suppression to further filter detections.

计算一个 ground truth 边界盒和 k 个先验框(Anchor)的交并比(IOU)值。 参数box: 元组或者数据，代表 ground truth 的长宽。 参数clusters: 形如(k,2)的numpy数组，其中k是聚类Anchor框的个数 返回：g

Log scalar variables.

Parses the data configuration file 解析数据配置文件、 返回的是字典 包括以下内容 classes= 1 # 类别数 train=data/custom/train.txt valid=data/c

Rescales bounding boxes to the original shape

Compute the average precision, given the recall and precision curves. Source: https://github.com/rafaelpadilla/Object-Detection-Metrics. # Ar

计算一个ground truth和k个Anchor的交并比的均值。

Compute the average precision, given the recall and precision curves. Code originally from https://github.com/rbgirshick/py-faster-rcnn. # A

Constructs module list of layer blocks from module configuration in module_defs 将配置文件的内容转成模型， 输入是一个列表，列表中是字典，每个字典代表一个网络块

Compute true positives, predicted scores and predicted labels per sample

利用IOU值进行K-means聚类 参数boxes: 形状为(r, 2)的ground truth框，其中r是ground truth的个数 参数k: Anchor的个数 参数dist: 距离函数 返回值：形状为(k, 2)的k个Anchor框

Parses the yolo-v3 layer configuration file and returns module definitions

Create a summary writer logging to log_dir.

list_path：数据集的路径，这个文件是一个 txt 文件，里面每一行是一个图片的路径

@:param path - path of the new weights file @:param cutoff - save layers between 0 and cutoff (cutoff = -1 -> all are sav

Log a scalar variable.

对权重进行初始化