def pre_bboxes(self, path1, path2=None):
files_path = glob.glob('{}/*jpg'.format(path1))
i = 0
total_best_bboxes = []
for image_path in files_path:
img = cv2.imread(image_path)
save_img_name = image_path[-5:-14:-1][::-1]
bboxes_pr = self.predict(img)
origial_bboxes = []
for pr_bboxes in bboxes_pr:
# print("pr_bboxes",pr_bboxes)
bboxes1 = pr_bboxes.tolist()
origial_bboxes.append(bboxes1)
origial_bboxes = torch.tensor(origial_bboxes)
# print("origial_bboxes",origial_bboxes,type(origial_bboxes))
best_bboxes = utils.py_nms(origial_bboxes, 0.1)
best_bboxes = best_bboxes.tolist()
save_img = utils.draw_bbox(img, best_bboxes, show_label=True)
if path2 != None:
save_image_path = os.path.join(path2, save_img_name + '.jpg')
cv2.imwrite(save_image_path, save_img)
print("保存成功")
best_bboxes.sort()
# print("best_bboxes",len(best_bboxes),best_bboxes)
total_best_bboxes.append(best_bboxes)
return total_best_bboxes
def two(self, path1, path2=None):
files_path = glob.glob('{}/*bmp'.format(path1))
for image_path in files_path:
img = cv2.imread(image_path)
save_img_name = image_path[-5:-14:-1][::-1]
H, W, C = img.shape
w1 = 832
m = 60
cw, epoch = math.modf(W / (w1 - m))
per_image = []
for i in range(int(epoch + 1)):
if i != int(W / (w1 - m)) and i == 0:
image = img[552:w1 + 552, w1 * i:w1 * (i + 1), :]
per_image.append(image)
elif i != int(W / (w1 - m)) and i >= 1:
image = img[552:w1 + 552,
w1 * i - m * i:w1 * (i + 1) - m * i, :]
per_image.append(image)
else:
image = img[552:w1 + 552, (w1 - m) * int(epoch):, :]
image1 = np.zeros((w1, w1, 3))
image1[:, :W - (w1 - m) * int(epoch), :] = image
per_image.append(image1.astype(np.float32))
total_bboxes_pr = []
for i, np_image in enumerate(per_image):
bboxes_pr = self.predict(np_image)
x_index = int(i)
if x_index == 0:
for boxes in bboxes_pr:
x1, y1, x2, y2, cls, _ = boxes
origial_x1 = x1
origial_y1 = y1 + 552
origial_x2 = x2
origial_y2 = y2 + 552
total_bboxes_pr.append(
np.array([
origial_x1, origial_y1, origial_x2, origial_y2,
cls, _
]))
else:
for boxes in bboxes_pr:
x1, y1, x2, y2, cls, _ = boxes
origial_x1 = x1 + (832 - 60) * x_index
origial_y1 = y1 + 552
origial_x2 = x2 + (832 - 60) * x_index
origial_y2 = y2 + 552
total_bboxes_pr.append(
np.array([
origial_x1, origial_y1, origial_x2, origial_y2,
cls, _
]))
origial_bboxes = []
for pr_bboxes in total_bboxes_pr:
# print("pr_bboxes",pr_bboxes)
bboxes1 = pr_bboxes.tolist()
origial_bboxes.append(bboxes1)
origial_bboxes = torch.tensor(origial_bboxes)
# print("origial_bboxes",origial_bboxes,type(origial_bboxes))
best_bboxes = utils.py_nms(origial_bboxes, 0.1)
print("best_bboxes", len(best_bboxes))
best_bboxes = best_bboxes.numpy().tolist()
origial_path = r'./docs/test_data_0610/'
origial_path_path = os.path.join(origial_path,
save_img_name + ".bmp")
origial_image = cv2.imread(origial_path_path)
big_origial_image = utils.draw_bbox(origial_image,
best_bboxes,
show_label=self.show_label)
if path2 != None:
save_image_path = os.path.join(path2,
save_img_name + ".pre" + '.bmp')
cv2.imwrite(save_image_path, big_origial_image)
print("保存成功")
exit()
def two(self, path1, path2=None):
lines = os.listdir(path1)
lines.sort(key=lambda x: int(x[:-4]))
total_per_name = []
total_index = []
total_cls = {}
for num, line in enumerate(lines):
annotation = line.split()
image_name = str(annotation).split('_')
image_names = image_name[0][2:]
index = image_name[1].split('.')[0]
total_index.append(int(index))
total_per_name.append(image_names)
per_name = list(set(total_per_name))
per_name.sort(key=total_per_name.index)
start_ind = 0
for i, per_image_name in enumerate(per_name):
end_ind = total_per_name.count(per_name[i]) + start_ind
total_cls[per_image_name] = total_index[start_ind:end_ind]
start_ind += end_ind
for i, (key, value) in enumerate(total_cls.items()):
total_bboxes_pr = []
for k in value:
image_path = os.path.join(path1,
key + "_" + "{}".format(k) + ".jpg")
image = cv2.imread(image_path)
bboxes_pr = self.predict(image)
x_index = int(k)
if x_index == 0:
for boxes in bboxes_pr:
x1, y1, x2, y2, cls, _ = boxes
origial_x1 = x1
origial_y1 = y1 + 552
origial_x2 = x2
origial_y2 = y2 + 552
total_bboxes_pr.append(
np.array([
origial_x1, origial_y1, origial_x2, origial_y2,
cls, _
]))
else:
for boxes in bboxes_pr:
x1, y1, x2, y2, cls, _ = boxes
origial_x1 = x1 + (832 - 60) * x_index
origial_y1 = y1 + 552
origial_x2 = x2 + (832 - 60) * x_index
origial_y2 = y2 + 552
total_bboxes_pr.append(
np.array([
origial_x1, origial_y1, origial_x2, origial_y2,
cls, _
]))
origial_bboxes = []
for pr_bboxes in total_bboxes_pr:
bboxes1 = pr_bboxes.tolist()
origial_bboxes.append(bboxes1)
origial_bboxes = torch.tensor(origial_bboxes)
best_bboxes = utils.py_nms(origial_bboxes, 0.1)
print("best_bboxes", len(best_bboxes))
best_bboxes = best_bboxes.numpy().tolist()
origial_path = r'./docs/trian_data_0610/'
origial_path_path = os.path.join(origial_path, key + ".bmp")
origial_image = cv2.imread(origial_path_path)
big_origial_image = utils.draw_bbox(origial_image,
best_bboxes,
show_label=self.show_label)
if path2 != None:
save_image_path = os.path.join(path2, key + '.bmp')
cv2.imwrite(save_image_path, big_origial_image)
print("保存成功")
def pre_bboxes(self, path1, path2=None, output_size=64):
files_path = glob.glob('{}/*bmp'.format(path1))
N = 0
lens = [201]
for image_path in files_path:
img = cv2.imread(image_path)
save_img_name = image_path[-5:-14:-1][::-1]
H, W, C = img.shape
w1 = 832
m = 60
cw, epoch = math.modf(W / (w1 - m))
per_image = []
for i in range(int(epoch + 1)):
if i != int(W / (w1 - m)) and i == 0:
image = img[552:w1 + 552, w1 * i:w1 * (i + 1), :]
per_image.append(image)
elif i != int(W / (w1 - m)) and i >= 1:
image = img[552:w1 + 552,
w1 * i - m * i:w1 * (i + 1) - m * i, :]
per_image.append(image)
else:
image = img[552:w1 + 552, (w1 - m) * int(epoch):, :]
image1 = np.zeros((w1, w1, 3))
image1[:, :W - (w1 - m) * int(epoch), :] = image
per_image.append(image1.astype(np.float32))
total_bboxes_pr = []
for i, np_image in enumerate(per_image):
bboxes_pr = self.predict(np_image)
x_index = int(i)
if x_index == 0:
for boxes in bboxes_pr:
x1, y1, x2, y2, cls, _ = boxes
origial_x1 = x1
origial_y1 = y1 + 552
origial_x2 = x2
origial_y2 = y2 + 552
total_bboxes_pr.append(
np.array([
origial_x1, origial_y1, origial_x2, origial_y2,
cls, _
]))
else:
for boxes in bboxes_pr:
x1, y1, x2, y2, cls, _ = boxes
origial_x1 = x1 + (832 - 60) * x_index
origial_y1 = y1 + 552
origial_x2 = x2 + (832 - 60) * x_index
origial_y2 = y2 + 552
total_bboxes_pr.append(
np.array([
origial_x1, origial_y1, origial_x2, origial_y2,
cls, _
]))
origial_bboxes = []
for pr_bboxes in total_bboxes_pr:
# print("pr_bboxes",pr_bboxes)
bboxes1 = pr_bboxes.tolist()
origial_bboxes.append(bboxes1)
origial_bboxes = torch.tensor(origial_bboxes)
# print("origial_bboxes",origial_bboxes,type(origial_bboxes))
best_bboxes = utils.py_nms(origial_bboxes, 0.2)
lens.append(len(best_bboxes) + lens[0])
print("best_bboxes", len(best_bboxes))
best_bboxes = best_bboxes.numpy().tolist()
best_bboxes.sort()
origial_path = r'./docs/trian_data_0610/'
origial_path_path = os.path.join(origial_path,
save_img_name + ".bmp")
origial_image = cv2.imread(origial_path_path)
predicted_dir_path = os.path.join(path2, save_img_name)
# print('predicted_dir_path', predicted_dir_path)
# if os.path.exists(predicted_dir_path): shutil.rmtree(predicted_dir_path)
os.mkdir(predicted_dir_path)
for kk, pre_bboxes in enumerate(best_bboxes):
Offset = []
cls_predicted_dir_path = os.path.join(predicted_dir_path,
str(kk + lens[N]))
os.mkdir(cls_predicted_dir_path)
for ii in range(6):
x1, y1, x2, y2, conf, cls = pre_bboxes
x1, y1, x2, y2 = int(x1), int(y1), int(x2), int(y2)
dw = x2 - x1
dh = y2 - y1
offset = random.randint(-5, 5)
Offset.append(offset)
print('Offset', Offset)
# if offset in Offset:
# continue
cx = int(x1 + dw / 2 + 1) + offset
cy = int(y1 + dh / 2 + 1) + offset
print("cx", cx, cy)
x11 = int(cx - output_size / 2)
x22 = x11 + output_size
y11 = int(cy - output_size / 2)
y22 = y11 + output_size
crop_image = origial_image[y11:y22, x11:x22]
save_image_path = os.path.join(
cls_predicted_dir_path,
str('{}'.format(ii)) + '.jpg')
# save_image_path = os.path.join(path2,save_img_name+"."+str(x11)+"."+str(y11)+"."+str(x22)+"."+str(y22)+ "." +str(round(conf,2))+"."+str(cls)+'.jpg')
cv2.imwrite(save_image_path, crop_image)
print("保存成功")
# print('一个循环结束')
N += 1
if N == 2:
break