def __getitem__(self, idx):
label = np.zeros((self.S, self.S, self.B * 5 + 13))
xml_path = self.annotation_path[idx]
img_path = self.images_path[idx]
xml_data = parse_xml(xml_path)
img_size = xml_data["size"]
image, ratio_x, ratio_y = self._image_processing(img_path, xml_data)
for i, ob in enumerate(xml_data["object"]):
object_class = np.zeros(13)
x, y, cell_x, cell_y, w, h = self._get_reat_axis(
ob["bndbox"], ratio_x, ratio_y)
# Each cell has B bounding box
b_id = self.kmeans.predict([[w, h]])[0]
label[cell_x, cell_y,
b_id * 5:(b_id + 1) * 5] = np.array([x, y, w, h, 1])
object_name = ob["name"]
object_class[self.classes.index(object_name)] = 1
label[cell_x, cell_y, -13:] = object_class
sample = {"image": np.array(image), "label": label}
if self.transform:
sample = self.transform(sample)
return sample
def calculate_anchor(input_size, train_annot_folder, saved_kmean_name):
annot_path_list = [
join(train_annot_folder, f) for f in listdir(train_annot_folder)
if isfile(join(train_annot_folder, f))
]
annot_path_list.sort()
kmean_list = []
for annot_name in annot_path_list:
xml_data = parse_xml(annot_name)
ratio_x, ratio_y = get_ratio(input_size, xml_data)
for ob in xml_data["object"]:
w, h = get_wh(input_size, ob["bndbox"], ratio_x, ratio_y)
kmean_list.append([w, h])
kmean_list = np.asarray(kmean_list)
kmeans = KMeans(n_clusters=5, random_state=0).fit(kmean_list)
pickle.dump(kmeans, open(saved_kmean_name, 'wb'))
def parse_annotation_jpeg(annotation_path, jpeg_path, gs):
"""
获取正负样本（注：忽略属性difficult为True的标注边界框）
正样本：候选建议与标注边界框IoU大于等于0.5
负样本：IoU大于0,小于0.5。为了进一步限制负样本数目，其大小必须大于标注框的1/5
"""
img = cv2.imread(jpeg_path)
selectivesearch.config(gs, img, strategy='q')
# 计算候选建议
rects = selectivesearch.get_rects(gs)
# 获取标注边界框
bndboxs = parse_xml(annotation_path)
# 标注框大小
maximum_bndbox_size = 0
for bndbox in bndboxs:
xmin, ymin, xmax, ymax = bndbox
bndbox_size = (ymax - ymin) * (xmax - xmin)
if bndbox_size > maximum_bndbox_size:
maximum_bndbox_size = bndbox_size
# 获取候选建议和标注边界框的IoU
iou_list = compute_ious(rects, bndboxs)
positive_list = list()
negative_list = list()
for i in range(len(iou_list)):
xmin, ymin, xmax, ymax = rects[i]
rect_size = (ymax - ymin) * (xmax - xmin)
iou_score = iou_list[i]
if iou_list[i] >= 0.5:
# 正样本
positive_list.append(rects[i])
if 0 < iou_list[i] < 0.5 and rect_size > maximum_bndbox_size / 5.0:
# 负样本
negative_list.append(rects[i])
else:
pass
return positive_list, negative_list
def parse_annotation_jpeg(annotation_path, jpeg_path, gs):
"""
获取正负样本（注：忽略属性difficult为True的标注边界框）
正样本：候选建议与标注边界框IoU大于等于0.5 + 标注边界框
负样本：IoU大于0.1,小于0.5
"""
img = cv2.imread(jpeg_path)
selectivesearch.config(gs, img, strategy='q')
# 计算候选建议
rects = selectivesearch.get_rects(gs)
# 获取标注边界框
bndboxs = parse_xml(annotation_path)
# 获取候选建议和标注边界框的IoU
iou_list = compute_ious(rects, bndboxs)
positive_list = list()
negative_list = list()
for i in range(len(iou_list)):
xmin, ymin, xmax, ymax = rects[i]
rect_size = (ymax - ymin) * (xmax - xmin)
iou_score = iou_list[i]
if iou_score >= 0.5:
# 正样本
positive_list.append(rects[i])
if 0.1 <= iou_score < 0.5:
# 负样本
negative_list.append(rects[i])
else:
pass
# 添加标注边界框到正样本列表
positive_list.extend(bndboxs)
return positive_list, negative_list
device = get_device()
transform = get_transform()
model = get_model(device=device)
# 创建selectivesearch对象
gs = selectivesearch.get_selective_search()
# test_img_path = '../imgs/000007.jpg'
# test_xml_path = '../imgs/000007.xml'
test_img_path = '../imgs/000012.jpg'
test_xml_path = '../imgs/000012.xml'
img = cv2.imread(test_img_path)
dst = copy.deepcopy(img)
bndboxs = util.parse_xml(test_xml_path)
for bndbox in bndboxs:
xmin, ymin, xmax, ymax = bndbox
cv2.rectangle(dst, (xmin, ymin), (xmax, ymax),
color=(0, 255, 0),
thickness=1)
# 候选区域建议
selectivesearch.config(gs, img, strategy='f')
rects = selectivesearch.get_rects(gs)
print('候选区域建议数目： %d' % len(rects))
# softmax = torch.softmax()
svm_thresh = 0.60
if __name__ == '__main__':
device = get_device()
transform = get_transform()
model = get_model(device=device)
# 创建selectivesearch对象
gs = selectivesearch.get_selective_search()
test_img_path = './data/voc_car/val/JPEGImages/000007.jpg'
test_xml_path = './data/voc_car/val/Annotations/000007.xml'
img = cv2.imread(test_img_path)
dst = copy.deepcopy(img)
bndboxs = parse_xml(test_xml_path)
for bndbox in bndboxs:
xmin, ymin, xmax, ymax = bndbox
cv2.rectangle(dst, (xmin, ymin), (xmax, ymax), color=(0, 255, 0), thickness=1)
# 候选区域建议
selectivesearch.config(gs, img, strategy='f')
rects = selectivesearch.get_rects(gs)
print('候选区域建议数目： %d' % len(rects))
# softmax = torch.softmax()
# 保存正样本边界框以及
score_list = list()
positive_list = list()
for rect in rects:
util.check_dir(dst_root_dir)
util.check_dir(dst_jpeg_dir)
util.check_dir(dst_bndbox_dir)
util.check_dir(dst_positive_dir)
samples = util.parse_car_csv(voc_car_train_dir)
res_samples = list()
total_positive_num = 0
for sample_name in samples:
# 提取正样本边界框坐标（IoU>=0.5）
positive_annotation_path = os.path.join(positive_annotation_dir, sample_name + '_1.csv')
positive_bndboxes = np.loadtxt(positive_annotation_path, dtype=np.int, delimiter=' ')
# 提取标注边界框
gt_annotation_path = os.path.join(gt_annotation_dir, sample_name + '.xml')
bndboxs = util.parse_xml(gt_annotation_path)
# 计算符合条件（IoU>0.6）的候选建议
positive_list = list()
if len(positive_bndboxes.shape) == 1 and len(positive_bndboxes) != 0:
scores = util.iou(positive_bndboxes, bndboxs)
if np.max(scores) > 0.6:
positive_list.append(positive_bndboxes)
elif len(positive_bndboxes.shape) == 2:
for positive_bndboxe in positive_bndboxes:
scores = util.iou(positive_bndboxe, bndboxs)
if np.max(scores) > 0.6:
positive_list.append(positive_bndboxe)
else:
pass
# 如果存在正样本边界框（IoU>0.6），那么保存相应的图片以及标注边界框