yolov3聚类自己数据的anchor box

前言

yolov3是一个很优秀的object-detection模型，其中的anchor box机制在多尺度检测上取得了不错的效果。然而，作者提供的anchor box值是基于voc和coco数据集上的，如果应用到自己数据集可能不完全适用，那么如何基于自己的训练数据聚类anchor box呢？好吧，源代码如下所示。

kemans.py

import numpy as npdef iou(box, clusters):"""Calculates the Intersection over Union (IoU) between a box and k clusters.:param box: tuple or array, shifted to the origin (i. e. width and height):param clusters: numpy array of shape (k, 2) where k is the number of clusters:return: numpy array of shape (k, 0) where k is the number of clusters"""x = np.minimum(clusters[:, 0], box[0])y = np.minimum(clusters[:, 1], box[1])if np.count_nonzero(x == 0) > 0 or np.count_nonzero(y == 0) > 0:raise ValueError("Box has no area")intersection = x * ybox_area = box[0] * box[1]cluster_area = clusters[:, 0] * clusters[:, 1]iou_ = intersection / (box_area + cluster_area - intersection)return iou_def avg_iou(boxes, clusters):"""Calculates the average Intersection over Union (IoU) between a numpy array of boxes and k clusters.:param boxes: numpy array of shape (r, 2), where r is the number of rows:param clusters: numpy array of shape (k, 2) where k is the number of clusters:return: average IoU as a single float"""return np.mean([np.max(iou(boxes[i], clusters)) for i in range(boxes.shape[0])])def translate_boxes(boxes):"""Translates all the boxes to the origin.:param boxes: numpy array of shape (r, 4):return: numpy array of shape (r, 2)"""new_boxes = boxes.copy()for row in range(new_boxes.shape[0]):new_boxes[row][2] = np.abs(new_boxes[row][2] - new_boxes[row][0])new_boxes[row][3] = np.abs(new_boxes[row][3] - new_boxes[row][1])return np.delete(new_boxes, [0, 1], axis=1)def kmeans(boxes, k, dist=np.median):"""Calculates k-means clustering with the Intersection over Union (IoU) metric.:param boxes: numpy array of shape (r, 2), where r is the number of rows:param k: number of clusters:param dist: distance function:return: numpy array of shape (k, 2)"""rows = boxes.shape[0]distances = np.empty((rows, k))last_clusters = np.zeros((rows,))np.random.seed()# the Forgy method will fail if the whole array contains the same rowsclusters = boxes[np.random.choice(rows, k, replace=False)]while True:for row in range(rows):distances[row] = 1 - iou(boxes[row], clusters)nearest_clusters = np.argmin(distances, axis=1)if (last_clusters == nearest_clusters).all():breakfor cluster in range(k):clusters[cluster] = dist(boxes[nearest_clusters == cluster], axis=0)last_clusters = nearest_clustersreturn clusters

cluster.py


import glob
import xml.etree.ElementTree as ET
import numpy as np
from kemans import kmeans, avg_iouANNOTATIONS_PATH = "F:/garbage/annotations"
CLUSTERS = 12def load_dataset(path):dataset = []for xml_file in glob.glob("{}/*xml".format(path)):tree = ET.parse(xml_file)height = int(tree.findtext("./size/height"))width = int(tree.findtext("./size/width"))for obj in tree.iter("object"):xmin = int(obj.findtext("bndbox/xmin")) / widthymin = int(obj.findtext("bndbox/ymin")) / heightxmax = int(obj.findtext("bndbox/xmax")) / widthymax = int(obj.findtext("bndbox/ymax")) / heightxmin = np.float64(xmin)ymin = np.float64(ymin)xmax = np.float64(xmax)ymax = np.float64(ymax)if xmax == xmin or ymax == ymin:print(xml_file)dataset.append([xmax - xmin, ymax - ymin])return np.array(dataset)if __name__ == '__main__':# print(__file__)data = load_dataset(ANNOTATIONS_PATH)out = kmeans(data, k=CLUSTERS)# clusters = [[10,13],[16,30],[33,23],[30,61],[62,45],[59,119],[116,90],[156,198],[373,326]]# out= np.array(clusters)/416.0print(out)print("Accuracy: {:.2f}%".format(avg_iou(data, out) * 100))print("Boxes:\n {}-{}".format(out[:, 0] * 608, out[:, 1] * 608))ratios = np.around(out[:, 0] / out[:, 1], decimals=2).tolist()print("Ratios:\n {}".format(sorted(ratios)))

最终聚类的结果：

[[0.3770724  0.15486111][0.03958333 0.0375    ][0.20677083 0.13686314][0.0703125  0.05347222][0.17916667 0.08472222][0.47375    0.24791667][0.08489583 0.08819444][0.125      0.0625    ][0.12447917 0.11319444][0.28541667 0.1025    ][0.2640625  0.21597222][0.14942708 0.1875    ]]
Accuracy: 75.99%
Boxes:[229.26001999  24.06666667 125.71666667  42.75       108.93333333288.04        51.61666667  76.          75.68333333 173.53333333160.55        90.85166667]-[ 94.15555556  22.8         83.21278721  32.51111111  51.51111111150.73333333  53.62222222  38.          68.82222222  62.32131.31111111 114.        ]
Ratios:[0.8, 0.96, 1.06, 1.1, 1.22, 1.31, 1.51, 1.91, 2.0, 2.11, 2.43, 2.78]

yolov3聚类自己数据的anchor box相关推荐

ECCV2020 AABO: Adaptive Anchor Box Optimization for Object Detection via Bayesian Sub-sampling论文翻译
ECCV2020 AABO论文翻译摘要 1.介绍 2.相关工作 3.提出的方法 3.1 初步分析 3.2 anchors的搜索空间优化 3.3 通过子抽样的贝叶斯锚优化 4.实验 4.1数据集,指标 ...
（Object detection）目标检测从入门到精通——第四部分anchor box
3.8 Anchor Boxes 到目前为止,对象检测中存在的一个问题是每个格子只能检测出一个对象,如果你想让一个格子检测出多个对象,你可以这么做,就是使用anchor box这个概念,我们从一个例子 ...
0927锚框（Anchor box）
锚框(Anchor box) 目标检测算法中,通常会在输入图像中采样大量的区域,然后判断这些区域中是否包含所感兴趣的目标,并调整区域边界从而更加准确地预测目标的真实边界框(ground-truth b ...
【YOLOv3从头训练数据篇】
YOLOv3从头训练数据篇前言数据下载数据可视化标签生成生成训练路径文件结语前言最近在忙着怎么从头实现YOLOv3,从网上找了很多教程,也在GitHub上面找到了挺多的代码的,有些能 ...
【目标检测】概念理解：region proposal、bounding box、anchor box、ground truth、IoU、NMS、RoI Pooling
最近刚接触图像识别,理解一些概念十分困难,尤其是动不动就冒出个看不懂的英语,让人抓狂.查了不少资料后做一个总结并加上一些自己的理解,理解若有误,烦请大家指出,相互学习. 本文主要对region pro ...
anchor box 是如何确定的?
文章目录 anchor box 是如何确定的? anchor box 框的位置是怎么设置的? anchor box 的数量和形状是怎么设置的? anchor box 的预测过程是怎样的? 1. 将 a ...
Anchor box的理解
Anchor box的理解这个概念最初是在Faster R-CNN中提出,此后在SSD.YOLOv2.YOLOv3等优秀的目标识别模型中得到了广泛的应用. 背景在Faster-RCNN中首次提出a ...
R语言ggplot2可视化可视化聚类图、使用geom_encircle函数绘制多边形标定属于同一聚类簇的数据点、并自定义每个聚类簇数据点的颜色、多边形框的颜色（Cluster Plot）、主副标题题注
R语言ggplot2可视化可视化聚类图.使用geom_encircle函数绘制多边形标定属于同一聚类簇的数据点.并自定义每个聚类簇数据点的颜色.多边形框的颜色(Cluster Plot).主副标题题注 ...
抛弃Anchor box和NMS，目标检测新范式开源：Sparse R-CNN
点击上方,选择星标或置顶,不定期资源大放送! 阅读大概需要10分钟 Follow小博主,每天更新前沿干货作者丨孙培泽@知乎来源丨https://zhuanlan.zhihu.com/p/31005 ...
计算机视觉滑动窗方法,图像分割相关技术之滑动窗口、RPN以及anchor box简介
图像分割相关技术之滑动窗口.RPN以及anchor box简介标签:## 时间:2019/11/17 11:07:25 作者:小木对象识别(object recognition)是计算机视觉(co ...

yolov3聚类自己数据的anchor box

前言

yolov3聚类自己数据的anchor box相关推荐

最新文章

热门文章