Yolo计算AP,mAP,生成P-R曲线
环境：Python3.7 Opencv Numpy Matplotlib
参考：
- https://blog.csdn.net/cgt19910923/article/details/80524173
- https://blog.csdn.net/amusi1994/article/details/81564504
- https://www.pianshen.com/article/59941009383/
前言：
-
在darknet版本中，用
darknet detector valid data cfg weight命令可以在result/目录下得到网络检测的输出。包括检测的图像名字、类别、概率、边界框位置（左上角和右下角）。 -
VOC数据集的xml格式标注文件在得到检测输出文件后用darknet/Script目录下的reval_voc_py3.py可以生成P-R曲线。
-
用Yolo_mark标注后的标注文件后缀为txt。
-
本文主要为修改reval_voc_py3.py文件后，利用yolo_mark的标注文件和valid命令生成的文件生成P-R曲线。
步骤如下：
- 修改voc_eval_py3.py中的parse_rec()函数。原函数是加载xml标签的，修改为加载txt标注文件的标签
- 修改VOC数据集参数传入。原reval_voc_py3需要传入VOC数据集的路径。
- reval_voc_py3 中，需要修改以下内容为自己的文件路径
label_path # label文件夹，标注txt和图像应在同一目录下
valid_file # valid命令生成的txt文件，在result/目录下。
name_path # name文件
output_dir # 生成的pkl保存路径
- 具体参考下列代码：
reval_voc_py.py:
#!/usr/bin/env python
import os, sys, argparse
import numpy as np
import _pickle as cPickle
from voc_eval_py3 import voc_eval
import matplotlib.pyplot as plt
def do_python_eval(label_path, valid_file, classes, output_dir = 'results'):
cachedir = os.path.join('./', 'annotations_cache')
aps = []
use_07_metric = False
print('VOC07 metric? ' + ('Yes' if use_07_metric else 'No'))
if not os.path.isdir(output_dir):
os.mkdir(output_dir)
for i, cls in enumerate(classes):
if cls == '__background__':
continue
rec, prec, ap = voc_eval(
label_path,
valid_file, cls, cachedir, ovthresh=0.5,
use_07_metric=use_07_metric)
aps += [ap]
print('AP for {} = {:.4f}'.format(cls, ap))
with open(os.path.join(output_dir, cls + '_pr.pkl'), 'wb') as f:
cPickle.dump({'rec': rec, 'prec': prec, 'ap': ap}, f)
print('Mean AP = {:.4f}'.format(np.mean(aps)))
print('~~~~~~~~')
print('Results:')
for ap in aps:
print('{:.3f}'.format(ap))
print('{:.3f}'.format(np.mean(aps)))
print('~~~~~~~~')
print('')
print('--------------------------------------------------------------')
print('Results computed with the **unofficial** Python eval code.')
print('Results should be very close to the official MATLAB eval code.')
print('-- Thanks, The Management')
print('--------------------------------------------------------------')
fr = open(cls + '_pr.pkl','rb')
inf = cPickle.load(fr)
fr.close()
x=inf['rec']
y=inf['prec']
plt.figure()
plt.xlabel('recall')
plt.ylabel('precision')
plt.title('PR cruve')
plt.plot(x,y)
plt.show()
print('AP：',inf['ap'])
if __name__ == '__main__':
label_path = r'E:\Datum\Data\dataset\vaild' #label文件夹
valid_file = r'8_10000_25.txt' #valid命令生成的txt文件，在result/目录下。
name_path = r'E:\Datum\Data\darknet\data\obj.name' #name文件
output_dir = os.path.abspath('./') #pkl保存路径
with open(name_path, 'r') as f:
lines = f.readlines()
classes = [t.strip('\n') for t in lines]
print('Evaluating detections')
do_python_eval(label_path, valid_file, classes, output_dir)
voc_ecal_py.py:
import xml.etree.ElementTree as ET
import os
import _pickle as cPickle
import numpy as np
import cv2
def parse_rec(label_path, label_name):
objects = []
label_file = os.path.join(label_path, label_name + '.txt')
img_file = os.path.join(label_path, label_name + '.jpg')
height, width, _ = cv2.imread(img_file).shape
with open(label_file) as f:
for line in f.readlines():
obj_struct = {}
obj_struct['name'] = 'car'
obj_struct['difficult'] = int(0)
center_x, center_y, width_b, height_b =[float(x) for x in line.split()[1:]]
obj_struct['bbox'] = [int(center_x * width - width * width_b / 2.0),
int(center_y * height - height * height_b / 2.0),
int(center_x * width + width * width_b / 2.0),
int(center_y * height + height * height_b / 2.0)]
objects.append(obj_struct)
return objects
def voc_ap(rec, prec, use_07_metric=False):
if use_07_metric:
# 11 point metric
ap = 0.
for t in np.arange(0., 1.1, 0.1):
if np.sum(rec >= t) == 0:
p = 0
else:
p = np.max(prec[rec >= t])
ap = ap + p / 11.
else:
# correct AP calculation
# first append sentinel values at the end
mrec = np.concatenate(([0.], rec, [1.]))
mpre = np.concatenate(([0.], prec, [0.]))
# compute the precision envelope
for i in range(mpre.size - 1, 0, -1):
mpre[i - 1] = np.maximum(mpre[i - 1], mpre[i])
# to calculate area under PR curve, look for points
# where X axis (recall) changes value
i = np.where(mrec[1:] != mrec[:-1])[0]
# and sum (\Delta recall) * prec
ap = np.sum((mrec[i + 1] - mrec[i]) * mpre[i + 1])
return ap
def voc_eval(label_path,
detpath,
classname,
cachedir,
ovthresh=0.5,
use_07_metric=False):
# first load gt
if not os.path.isdir(cachedir):
os.mkdir(cachedir)
cachefile = os.path.join(cachedir, 'annots.pkl')
label_file = []
for f in os.listdir(label_path):
file, tmp = f.split('.')
if tmp == 'txt':
label_file.append(file)
if not os.path.isfile(cachefile):
# load annots
recs = {}
for label_name in label_file:
recs[label_name] = parse_rec(label_path, label_name)
with open(cachefile, 'wb') as f:
cPickle.dump(recs, f)
else:
# load
print('!!! cachefile = ',cachefile)
with open(cachefile, 'rb') as f:
recs = cPickle.load(f)
# extract gt objects for this class
class_recs = {}
npos = 0
for label_name in label_file:
# for imagename in imagenames:
R = [obj for obj in recs[label_name] if obj['name'] == classname]
bbox = np.array([x['bbox'] for x in R])
difficult = np.array([x['difficult'] for x in R]).astype(np.bool)
det = [False] * len(R)
npos = npos + sum(~difficult)
class_recs[label_name] = {'bbox': bbox,
'difficult': difficult,
'det': det}
# read dets
detfile = detpath
# detfile = detpath.format(classname)
with open(detfile, 'r') as f:
lines = f.readlines()
splitlines = [x.strip().split(' ') for x in lines]
image_ids = [x[0] for x in splitlines]
confidence = np.array([float(x[1]) for x in splitlines])
BB = np.array([[float(z) for z in x[2:]] for x in splitlines])
# sort by confidence
sorted_ind = np.argsort(-confidence)
sorted_scores = np.sort(-confidence)
BB = BB[sorted_ind, :]
image_ids = [image_ids[x] for x in sorted_ind]
# go down dets and mark TPs and FPs
nd = len(image_ids)
# print(image_ids)
# print(nd)
tp = np.zeros(nd)
fp = np.zeros(nd)
for d in range(nd):
R = class_recs[image_ids[d]]
bb = BB[d, :].astype(float)
ovmax = -np.inf
BBGT = R['bbox'].astype(float)
if BBGT.size > 0:
# compute overlaps
# intersection
ixmin = np.maximum(BBGT[:, 0], bb[0])
iymin = np.maximum(BBGT[:, 1], bb[1])
ixmax = np.minimum(BBGT[:, 2], bb[2])
iymax = np.minimum(BBGT[:, 3], bb[3])
iw = np.maximum(ixmax - ixmin + 1., 0.)
ih = np.maximum(iymax - iymin + 1., 0.)
inters = iw * ih
# union
uni = ((bb[2] - bb[0] + 1.) * (bb[3] - bb[1] + 1.) +
(BBGT[:, 2] - BBGT[:, 0] + 1.) *
(BBGT[:, 3] - BBGT[:, 1] + 1.) - inters)
overlaps = inters / uni
ovmax = np.max(overlaps)
jmax = np.argmax(overlaps)
if ovmax > ovthresh:
if not R['difficult'][jmax]:
if not R['det'][jmax]:
tp[d] = 1.
R['det'][jmax] = 1
else:
fp[d] = 1.
else:
fp[d] = 1.
# compute precision recall
fp = np.cumsum(fp)
tp = np.cumsum(tp)
rec = tp / float(npos)
# avoid divide by zero in case the first detection matches a difficult
# ground truth
prec = tp / np.maximum(tp + fp, np.finfo(np.float64).eps)
ap = voc_ap(rec, prec, use_07_metric)
return rec, prec, ap