目标检测的mAP值的计算(python实现)
目标检测的mAP值的计算(python实现)（针对VOC数据集）
准备条件
- 训练好的目标检测模型
- voc数据集，将测试集图片放入images文件夹中
- 将测试集图片对应的xml文件放入annotations文件夹中
1. 导入模块
import os
import re
import cv2
import glob
import json
import math
import shutil
import numpy as np
import tensorflow as tf
from PIL import Image2. 导入训练好的模型
PATH_TO_CKPT = 'model/frozen_inference_graph.pb'
detection_graph = tf.Graph()
with detection_graph.as_default():
od_graph_def = tf.GraphDef()
with tf.gfile.GFile(PATH_TO_CKPT, 'rb') as fid:
serialized_graph = fid.read()
od_graph_def.ParseFromString(serialized_graph)
tf.import_graph_def(od_graph_def, name='')
with detection_graph.as_default():
with tf.Session(graph=detection_graph) as sess:
image_tensor = detection_graph.get_tensor_by_name('image_tensor:0')
detection_boxes = detection_graph.get_tensor_by_name('detection_boxes:0')
detection_scores = detection_graph.get_tensor_by_name('detection_scores:0')
detection_classes = detection_graph.get_tensor_by_name('detection_classes:0')
num_detections = detection_graph.get_tensor_by_name('num_detections:0')3. 使用训练好的模型对测试图片进行预测，生成预测文件
PATH_TO_IMAGES = 'images/'
PREDICTED_PATH = 'predicted/'
if not os.path.exists(PREDICTED_PATH):
os.makedirs(PREDICTED_PATH)
for filename in os.listdir(PATH_TO_IMAGES):
if filename.lower().endswith(('.bmp', '.dib', '.png', '.jpg', '.jpeg', '.pbm', '.pgm', '.ppm', '.tif', '.tiff')):
image_name = PATH_TO_IMAGES + filename
image = cv2.imread(image_name)
image_np = np.array(image).astype(np.uint8)
image_np_expanded = np.expand_dims(image_np, axis=0)
(boxes, scores, classes, num) = sess.run(
[detection_boxes, detection_scores, detection_classes, num_detections],
feed_dict={
image_tensor: image_np_expanded})
img = cv2.cvtColor(np.array(image_np), cv2.COLOR_RGB2BGR)
shape = img.shape
p1 = (int(boxes[0][0][0] * shape[0]), int(boxes[0][0][1] * shape[1]))
p2 = (int(boxes[0][0][2] * shape[0]), int(boxes[0][0][3] * shape[1]))
with open(PREDICTED_PATH + "2007_" + filename[:-4] + ".txt","w") as f:
f.write("instrument"+" "+str(round(scores[0][0],6))+" "+str(p1[::-1][0])+" "+str(p1[::-1][1])+" "+str(p2[::-1][0])+" "+str(p2[::-1][1])+"\n")4. 对已标记的xml文件进行处理，生成真实标签文件
PATH_TO_XML = 'annotations/'
GROUND_TRUTH_PATH = 'ground-truth/'
if not os.path.exists(GROUND_TRUTH_PATH):
os.makedirs(GROUND_TRUTH_PATH)
for filename in os.listdir(PATH_TO_XML):
if filename.lower().endswith(('.xml')):
xml_name = os.path.join(PATH_TO_XML, filename)
with open(xml_name,"r") as f:
f1 = open(GROUND_TRUTH_PATH + "2007_" + filename[:-4] + ".txt","w")
for line in f.readlines():
classes = re.match(r".*(.*) .*",line,re.I)
xmin = re.match(r".*(.*) .*",line,re.I)
ymin = re.match(r".*(.*) .*",line,re.I)
xmax = re.match(r".*(.*) .*",line,re.I)
ymax = re.match(r".*(.*) .*",line,re.I)
if classes:
f1.write(classes.group(1) + " ")
if xmin:
f1.write(xmin.group(1) + " ")
if ymin:
f1.write(ymin.group(1) + " ")
if xmax:
f1.write(xmax.group(1) + " ")
if ymax:
f1.write(ymax.group(1) + "\n")
f1.close()
f.close()5. 给定召回率和精度数组，计算AP
def voc_ap(rec, prec):
rec.insert(0, 0.0)
rec.append(1.0)
mrec = rec[:]
prec.insert(0, 0.0)
prec.append(0.0)
mpre = prec[:]
for i in range(len(mpre)-2, -1, -1):
mpre[i] = max(mpre[i], mpre[i+1])
i_list = []
for i in range(1, len(mrec)):
if mrec[i] != mrec[i-1]:
i_list.append(i)
ap = 0.0
for i in i_list:
ap += ((mrec[i]-mrec[i-1])*mpre[i])
return ap6. 处理真实标签文件，创建列表存储真实标签数据存在的所有类的数据
TEMP_FILES_PATH = ".temp_files"
if not os.path.exists(TEMP_FILES_PATH):
os.makedirs(TEMP_FILES_PATH)
ground_truth_files_list = glob.glob('ground-truth/*.txt')
ground_truth_files_list.sort()
gt_counter_per_class = {
}
counter_images_per_class = {
}
for txt_file in ground_truth_files_list:
file_id = txt_file.split(".txt",1)[0]
file_id = os.path.basename(os.path.normpath(file_id))
f = open(txt_file)
lines_list = f.readlines()
f.close()
lines_list = [x.strip() for x in lines_list]
bounding_boxes = []
already_seen_classes = []
for line in lines_list:
class_name, left, top, right, bottom = line.split()
bbox = left + " " + top + " " + right + " " +bottom
bounding_boxes.append({
"class_name":class_name, "bbox":bbox, "used":False})
if class_name in gt_counter_per_class:
gt_counter_per_class[class_name] += 1
else:
gt_counter_per_class[class_name] = 1
if class_name not in already_seen_classes:
if class_name in counter_images_per_class:
counter_images_per_class[class_name] += 1
else:
counter_images_per_class[class_name] = 1
already_seen_classes.append(class_name)
# 将每个真实标签数据写入一个临时的.json文件中
with open(TEMP_FILES_PATH + "/" + file_id + "_ground_truth.json", 'w') as outfile:
json.dump(bounding_boxes, outfile)
gt_classes = list(gt_counter_per_class.keys())
gt_classes = sorted(gt_classes)
n_classes = len(gt_classes)7. 处理预测文件，将每个预测数据写入一个临时的.json文件中
predicted_files_list = glob.glob('predicted/*.txt')
predicted_files_list.sort()
for class_index, class_name in enumerate(gt_classes):
bounding_boxes = []
for txt_file in predicted_files_list:
file_id = txt_file.split(".txt",1)[0]
file_id = os.path.basename(os.path.normpath(file_id))
f = open(txt_file)
lines = f.readlines()
f.close()
lines = [x.strip() for x in lines]
for line in lines:
tmp_class_name, confidence, left, top, right, bottom = line.split()
if tmp_class_name == class_name:
bbox = left + " " + top + " " + right + " " +bottom
bounding_boxes.append({
"confidence":confidence, "file_id":file_id, "bbox":bbox})
bounding_boxes.sort(key=lambda x:float(x['confidence']), reverse=True)
with open(TEMP_FILES_PATH + "/" + class_name + "_predictions.json", 'w') as outfile:
json.dump(bounding_boxes, outfile)8. 设置iou阈值（默认0.5），计算mAP，结果写入results.txt文件中
MINOVERLAP = 0.5 # iou阈值
sum_AP = 0.0
results_files_path = "results"
if os.path.exists(results_files_path):
shutil.rmtree(results_files_path)
os.makedirs(results_files_path)
with open(results_files_path + "/results.txt", 'w') as results_file:
results_file.write("# AP and precision/recall per class\n")
count_true_positives = {
}
for class_index, class_name in enumerate(gt_classes):
count_true_positives[class_name] = 0
#导入预测数据
predictions_file = TEMP_FILES_PATH + "/" + class_name + "_predictions.json"
predictions_data = json.load(open(predictions_file))
nd = len(predictions_data)
tp = [0] * nd
fp = [0] * nd
for idx, prediction in enumerate(predictions_data):
file_id = prediction["file_id"]
gt_file = TEMP_FILES_PATH + "/" + file_id + "_ground_truth.json"
ground_truth_data = json.load(open(gt_file))
ovmax = -1
gt_match = -1
# 导入预测框
bb = [ float(x) for x in prediction["bbox"].split() ]
for obj in ground_truth_data:
if obj["class_name"] == class_name:
bbgt = [ float(x) for x in obj["bbox"].split() ]
bi = [max(bb[0],bbgt[0]), max(bb[1],bbgt[1]), min(bb[2],bbgt[2]), min(bb[3],bbgt[3])]
iw = bi[2] - bi[0] + 1
ih = bi[3] - bi[1] + 1
if iw > 0 and ih > 0:
# 计算IOU
ua = (bb[2] - bb[0] + 1) * (bb[3] - bb[1] + 1) + (bbgt[2] - bbgt[0]
+ 1) * (bbgt[3] - bbgt[1] + 1) - iw * ih
ov = iw * ih / ua
if ov > ovmax:
ovmax = ov
gt_match = obj
min_overlap = MINOVERLAP
if ovmax >= min_overlap:
if not bool(gt_match["used"]):
tp[idx] = 1
gt_match["used"] = True
count_true_positives[class_name] += 1
with open(gt_file, 'w') as f:
f.write(json.dumps(ground_truth_data))
else:
fp[idx] = 1
else:
fp[idx] = 1
if ovmax > 0:
status = "INSUFFICIENT OVERLAP"
# 计算召回率和精度
cumsum = 0
for idx, val in enumerate(fp):
fp[idx] += cumsum
cumsum += val
cumsum = 0
for idx, val in enumerate(tp):
tp[idx] += cumsum
cumsum += val
rec = tp[:]
for idx, val in enumerate(tp):
rec[idx] = float(tp[idx]) / gt_counter_per_class[class_name]
prec = tp[:]
for idx, val in enumerate(tp):
prec[idx] = float(tp[idx]) / (fp[idx] + tp[idx])
# 计算AP
ap = voc_ap(rec[:], prec[:])
sum_AP += ap
text = "{0:.2f}%".format(ap*100) + " = " + class_name + " AP "
# 结果写入results.txt文件中
rounded_prec = [ '%.2f' % elem for elem in prec ]
rounded_rec = [ '%.2f' % elem for elem in rec ]
results_file.write(text + "\n Precision: " + str(rounded_prec) + "\n Recall :" + str(rounded_rec) + "\n\n")
n_images = counter_images_per_class[class_name]
results_file.write("\n# mAP of all classes\n")
# 计算mAP
mAP = sum_AP / n_classes
text = "mAP = {0:.2f}%".format(mAP*100)
results_file.write(text + "\n")
shutil.rmtree(TEMP_FILES_PATH)
9. 计算预测总数，将各个类别的总数写入results.txt
pred_counter_per_class = {
}
for txt_file in predicted_files_list:
f = open(txt_file)
lines_list = f.readlines()
f.close()
lines_list = [x.strip() for x in lines]
for line in lines_list:
class_name = line.split()[0]
if class_name in pred_counter_per_class:
pred_counter_per_class[class_name] += 1
else:
pred_counter_per_class[class_name] = 1
pred_classes = list(pred_counter_per_class.keys())
with open(results_files_path + "/results.txt", 'a') as results_file:
results_file.write("\n# Number of ground-truth objects per class\n")
for class_name in sorted(gt_counter_per_class):
results_file.write(class_name + ": " + str(gt_counter_per_class[class_name]) + "\n")
10. 将各个类别的TP/FP数量写入results.txt
for class_name in pred_classes:
if class_name not in gt_classes:
count_true_positives[class_name] = 0
with open(results_files_path + "/results.txt", 'a') as results_file:
results_file.write("\n# Number of predicted objects per class\n")
for class_name in sorted(pred_classes):
n_pred = pred_counter_per_class[class_name]
text = class_name + ": " + str(n_pred)
text += " (tp:" + str(count_true_positives[class_name]) + ""
text += ", fp:" + str(n_pred - count_true_positives[class_name]) + ")\n"
results_file.write(text)
results_file.close()
代码文件下载地址