def extract_detections(frame, min_confidence=0.6, labels=None):
"""Extract detections from frame.
frame: numpy array WxHx3
returns: numpy int array Cx5 [[label_id, xmin, ymin, xmax, ymax]]
"""
# Write code here
# First, convert frame to float and resize to 300x300
w, h = np.shape(frame)[:2]
frame = resize(frame.astype(float), (300, 300, 3))
# Then use preprocess_input, model.predict and bbox_util.detection_out
# Use help(...) function to help
x = preprocess_input(frame)
x = np.expand_dims(x, axis=0)
y = model.predict(x)
results = bbox_util.detection_out(y)[0]
# Select detections with confidence > min_confidence
results = results[results[:, 1] > min_confidence]
# If label set is known, use it
if labels is not None:
result_labels = results[:, 0].astype(np.int32)
indeces = [
i for i, l in enumerate(result_labels)
if VOC_CLASSES[l - 1] in labels
]
results = results[indeces]
else:
results[:, 0] = -1
# Remove confidence column from result
results = np.delete(results, 1, 1)
# Resize detection coords to input image shape.
# Didn't you forget to save it before resize?
results[:, 1] *= h
results[:, 2] *= w
results[:, 3] *= h
results[:, 4] *= w
# Return result
return detection_cast(results.astype(int))
def predict_result(model, file_path):
# print("+++++++++++++++++++++++++++++++++")
# print(file_path)
global cnt
img_path = file_path
img = image.load_img(img_path, target_size=(224, 224))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
# print(x)
#print('Input image shape:', x.shape)
#im = Image.open(file_path)
#im.show()
preds = model.predict(x)
#print('Predicted:', decode_predictions(preds))
label = decode_predictions(preds)[0][0][1]
if label == 'toilet_tissue':
print('Predicted:', decode_predictions(preds))
print(file_path)
cnt += 1
else:
print("+++++++++++++++++++++++++++++++++++++++")
print('Predicted:', decode_predictions(preds))
def model_predict(img_path, model):
img = image.load_img(img_path, target_size=(150, 150))
# Preprocessing the image
x = image.img_to_array(img)
# x = np.true_divide(x, 255)
# Scaling
# x = x / 255
x = np.expand_dims(x, axis=0)
# Be careful how your trained model deals with the input
# otherwise, it won't make correct prediction!
x = preprocess_input(x)
preds = model.predict(x)
# preds = np.argmax(preds, axis=0)
if preds[0][0] == 0:
preds = "This is Defective Insulator"
else:
preds = "This is Normal Insulator"
return preds
def model_predict(img_path):
img = image.load_img(img_path, target_size=(200, 200))
model = load_model('MobileNet_model.h5')
# Preprocessing the image
x = image.img_to_array(img)
# x = np.true_divide(x, 255)
## Scaling
x = x / 255
x = np.expand_dims(x, axis=0)
# Be careful how your trained model deals with the input
# otherwise, it won't make correct prediction!
x = preprocess_input(x)
preds = model.predict(x)
preds = np.argmax(preds, axis=1)
if preds == 0:
preds = "The Person is not Infected With Pneumonia"
else:
preds = "The Person is Infected With Pneumonia"
return preds
def model_predict(img_path, model):
img = image.load_img(img_path, target_size=(224, 224))
x = image.img_to_array(img)
x = x / 255
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
preds = model_predict(x)
preds = np.argmax(preds, axis=1)
if preds == 0:
preds = "The Car is Audi !!!"
elif preds == 1:
preds = "The Car is Lamborghini !!!"
else:
preds = "The Car is Mercedes !!!"
return preds
def model_predict(img_path, model):
img = image.load_img(img_path, target_size=(224, 224))
# Preprocessing the image
x = image.img_to_array(img)
# x = np.true_divide(x, 255)
## Scaling
x = x / 255
x = np.expand_dims(x, axis=0)
# Be careful how your trained model deals with the input
# otherwise, it won't make correct prediction!
x = preprocess_input(x)
preds = model.predict(x)
preds = np.argmax(preds, axis=1)
if preds == 0:
preds = "Normal"
else:
preds = "Covid"
return preds
def prepare_train_data(vgg_face):
# Prepare Train Data
x_train = []
y_train = []
person_rep = dict()
person_folders = os.listdir('./faces/')
for i, person in enumerate(person_folders):
print("preparing train set for person: " + person)
person_rep[i] = person
image_names = os.listdir(r'.\faces\\' + person + '\\')
for image_name in image_names:
print("\tphoto: " + image_name)
img = load_img('./faces/' + person + '/' + image_name,
target_size=(224, 224))
img = img_to_array(img)
img = np.expand_dims(img, axis=0)
img = preprocess_input(img)
img_encode = vgg_face(img)
x_train.append(np.squeeze(K.eval(img_encode)).tolist())
y_train.append(i)
return person_rep, np.array(x_train), np.array(y_train)
def model_predict(img_path, model):
img = image.load_img(img_path, target_size=(224, 224))
# Preprocessing the image
x = image.img_to_array(img)
x = np.true_divide(x, 255)
## Scaling
x=x/255
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
preds = model.predict(x)
preds=np.argmax(preds, axis=1)
if preds==0:
preds="The Car IS Audi"
elif preds==1:
preds="The Car is Lamborghini"
else:
preds="The Car Is Mercedes"
return preds
def get_integrated_gradients(model,
img_input,
top_pred_idx,
baseline=None,
num_steps=50):
if baseline is None:
baseline = np.zeros(img_size).astype(np.float32)
else:
baseline = baseline.astype(np.float32)
# 1. Do interpolation.
img_input = img_input.astype(np.float32)
interpolated_image = [
baseline + (step / num_steps) * (img_input - baseline)
for step in range(num_steps + 1)
]
interpolated_image = np.array(interpolated_image).astype(np.float32)
# 2. Preprocess the interpolated images
# interpolated_image = xception.preprocess_input(interpolated_image)
interpolated_image = preprocess_input(interpolated_image, mode="tf")
# 3. Get the gradients
grads = []
for i, img in enumerate(interpolated_image):
img = tf.expand_dims(img, axis=0)
grad = compute_gradients(model, img, top_pred_idx=top_pred_idx)
grads.append(grad[0])
grads = tf.convert_to_tensor(grads, dtype=tf.float32)
# 4. Approximate the integral using the trapezoidal rule
grads = (grads[:-1] + grads[1:]) / 2.0
avg_grads = tf.reduce_mean(grads, axis=0)
# 5. Calculate integrated gradients and return
integrated_grads = (img_input - baseline) * avg_grads
return integrated_grads
def generate(self, train=True):
while True:
if train:
# 打乱
shuffle(self.train_lines)
lines = self.train_lines
else:
shuffle(self.val_lines)
lines = self.val_lines
inputs = []
targets = []
for annotation_line in lines:
img,y=self.get_random_data(annotation_line,self.image_size[0:2])
if len(y)!=0:
boxes = np.array(y[:,:4],dtype=np.float32)
boxes[:,0] = boxes[:,0]/self.image_size[1]
boxes[:,1] = boxes[:,1]/self.image_size[0]
boxes[:,2] = boxes[:,2]/self.image_size[1]
boxes[:,3] = boxes[:,3]/self.image_size[0]
one_hot_label = np.eye(self.num_classes)[np.array(y[:,4],np.int32)]
if ((boxes[:,3]-boxes[:,1])<=0).any() and ((boxes[:,2]-boxes[:,0])<=0).any():
continue
y = np.concatenate([boxes,one_hot_label],axis=-1)
# print('y', tf.shape(y))
y = self.bbox_util.assign_boxes(y)
# print('y', tf.shape(y))
inputs.append(img)
targets.append(y)
if len(targets) == self.batch_size:
tmp_inp = np.array(inputs)
tmp_targets = np.array(targets)
inputs = []
targets = []
# print(tf.shape(preprocess_input(tmp_inp)), tf.shape(tmp_targets))
yield preprocess_input(tmp_inp), tmp_targets
def predict_with_gradient(dirs):
'''
Classify image then create heatmap (gradient). Save combination of image, gradient and gradient on image.
:param dirs: list of paths to all data items
:return: nothing
'''
# loop through all photos in dirs and predict
for item in dirs:
orig = cv2.imread(path + item)
image = load_img(path + item)
image = img_to_array(image)
image = np.expand_dims(image, axis=0)
image = imagenet_utils.preprocess_input(image)
# use the network to make predictions
preds = model.predict(image)
i = np.argmax(preds[0])
label = "Gradient"
# initialize our gradient class activation map and build the heatmap
cam = GradCAM(model, i)
heatmap = cam.compute_heatmap(image)
# resize the resulting heatmap to the original input image dimensions
# and then overlay heatmap on top of the image
heatmap = cv2.resize(heatmap, (orig.shape[1], orig.shape[0]))
(heatmap, output) = cam.overlay_heatmap(heatmap, orig, alpha=0.5)
# draw the predicted label on the output image
cv2.rectangle(output, (0, 0), (340, 40), (0, 0, 0), -1)
cv2.putText(output, label, (10, 25), cv2.FONT_HERSHEY_SIMPLEX, 0.8,
(255, 255, 255), 2)
# display the original image and resulting heatmap and output image
# to our screen
output = np.vstack([orig, heatmap, output])
output = imutils.resize(output, height=700)
# cv2.imshow("Output", output)
cv2.waitKey(0)
cv2.imwrite(my_path_results_pos + item, output)
def recognize(img, outputs, class_names, vgg_face):
person_rep = dict()
person_names = ["angelamerkel", "jinping", "trump"]
for person in person_names:
embed = np.loadtxt(person + ".txt")
person_rep[person] = embed
boxes, objectness, classes, nums = outputs
boxes, objectness, classes, nums = boxes[0], objectness[0], classes[
0], nums[0]
wh = np.flip(img.shape[0:2])
for i in range(nums):
x1y1 = tuple((np.array(boxes[i][0:2]) * wh).astype(np.int32))
x2y2 = tuple((np.array(boxes[i][2:4]) * wh).astype(np.int32))
if class_names[int(classes[i])] == "face":
img_crop = img[x1y1[1]:x2y2[1], x1y1[0]:x2y2[0]]
crop_img = img_to_array(img_crop)
crop_img = np.expand_dims(crop_img, axis=0)
crop_img = preprocess_input(crop_img)
img_encode = vgg_face(transform_images(crop_img, 224))
embed = K.eval(img_encode)
name, score = get_match(person_rep, embed, 0.3)
img = cv2.rectangle(img, x1y1, x2y2, (205, 0, 0), 2)
img = cv2.putText(img, '{} {:.4f}'.format(name, score), x1y1,
cv2.FONT_HERSHEY_COMPLEX_SMALL, 1, (255, 0, 255),
2)
else:
img = cv2.rectangle(img, x1y1, x2y2, (255, 0, 0), 2)
img = cv2.putText(
img, '{} {:.4f}'.format(class_names[int(classes[i])],
objectness[i]), x1y1,
cv2.FONT_HERSHEY_COMPLEX_SMALL, 1, (0, 0, 255), 2)
return img
def main():
parser = argparser()
args = parser.parse_args()
image_path = args.image
layer_name = args.layer_name
feature_to_visualize = args.feature
visualize_mode = args.mode
model = vgg16.VGG16(weights='imagenet', include_top=True)
layer_dict = dict([(layer.name, layer) for layer in model.layers])
if not layer_name in layer_dict:
print('Wrong layer name')
sys.exit()
# Load data and preprocess
img = Image.open(image_path)
img = img.resize((224, 224))
img_array = np.array(img)
img_array = img_array[np.newaxis, :]
img_array = img_array.astype(np.float)
img_array = imagenet_utils.preprocess_input(img_array)
deconv = visualize(model, img_array, layer_name, feature_to_visualize,
visualize_mode)
# postprocess and save image
deconv = deconv - deconv.min()
deconv *= 1.0 / (deconv.max() + 1e-8)
deconv = deconv[:, :, ::-1]
uint8_deconv = (deconv * 255).astype(np.uint8)
img = Image.fromarray(uint8_deconv, 'RGB')
img.save('results/{}_{}_{}.png'.format(layer_name, feature_to_visualize,
visualize_mode))
print('Saved: results/{}_{}_{}.png'.format(layer_name,
feature_to_visualize,
visualize_mode))
def get_model_in_blocks(self, model_function, include_top=True):
# Load the vanilla model
model = model_function(pretrained=True)
# Instantiate consecutive blocks
blocks = []
# Add preprocess_input
blocks.append(
Lambda(
lambda x: preprocess_input(x, mode=self.preprocess_input_mode),
name="preprocess_input"))
# Discard the last pooling layer
blocks += model.features.children[:-1]
if include_top:
# Add a GlobalAveragePooling2D layer
blocks.append(GlobalAveragePooling2D())
# Add the dense layer with softmax
blocks += [model.output1, Softmax()]
return blocks
def recognize(self,draw):
height,width,_ = np.shape(draw)
draw_rgb = cv.cvtColor(draw,cv.COLOR_BGR2RGB)
rectangles = self.mtcnn_model.detectFace(draw_rgb,self.threshold)
if len(rectangles)==0:
return
rectangles = np.array(rectangles,dtype=np.int32)
rectangles[:,0] = np.clip(rectangles[:,0],0,width)
rectangles[:,1] = np.clip(rectangles[:,1],0,height)
rectangles[:,2] = np.clip(rectangles[:,2],0,width)
rectangles[:,3] = np.clip(rectangles[:,3],0,height)
rectangles_temp = utils.rect2square(np.array(rectangles,dtype=np.int32))
rectangles_temp[:,0] = np.clip(rectangles_temp[:,0],0,width)
rectangles_temp[:,1] = np.clip(rectangles_temp[:,1],0,height)
rectangles_temp[:,2] = np.clip(rectangles_temp[:,2],0,width)
rectangles_temp[:,3] = np.clip(rectangles_temp[:,3],0,height)
for rectangle in rectangles_temp:
# 获取landmark在小图中的坐标
landmark = (np.reshape(rectangle[5:15],(5,2)) - np.array([int(rectangle[0]),int(rectangle[1])]))/(rectangle[3]-rectangle[1])*160
# 截取图像
crop_img = draw_rgb[int(rectangle[1]):int(rectangle[3]), int(rectangle[0]):int(rectangle[2])]
crop_img = cv.resize(crop_img,(self.Crop_HEIGHT,self.Crop_WIDTH))
# 对齐
new_img,_ = utils.Alignment_1(crop_img,landmark)
# 归一化
new_img = preprocess_input(np.reshape(np.array(new_img,np.float64),[1,self.Crop_HEIGHT,self.Crop_WIDTH,3]))
rectangles = rectangles[:,0:4]
#draw = cv2.cvtColor(draw, cv2.COLOR_RGB2BGR)
for (left, top, right, bottom) in rectangles:
cv.rectangle(draw, (left, top), (right, bottom), (255, 0, 0), 2)
return draw
def generate(self, train=True):
while True:
if train:
shuffle(self.train_keys)
keys = self.train_keys
else:
shuffle(self.val_keys)
keys = self.val_keys
inputs = []
targets = []
for key in keys:
img_path = self.path_prefix + key
img = imread(img_path).astype('float32')
y = self.gt[key].copy()
if train and self.do_crop:
img, y = self.random_sized_crop(img, y)
img = imresize(img, self.image_size).astype('float32')
if train:
shuffle(self.color_jitter)
for jitter in self.color_jitter:
img = jitter(img)
if self.lighting_std:
img = self.lighting(img)
if self.hflip_prob > 0:
img, y = self.horizontal_flip(img, y)
if self.vflip_prob > 0:
img, y = self.vertical_flip(img, y)
y = self.bbox_util.assign_boxes(y)
inputs.append(img)
targets.append(y)
if len(targets) == self.batch_size:
tmp_inp = np.array(inputs)
tmp_targets = np.array(targets)
inputs = []
targets = []
yield preprocess_input(tmp_inp), tmp_targets
def load_paired_img_wrd(folder):
class_names = [fold for fold in os.listdir(folder) if ".DS" not in fold]
image_list = []
labels_list = []
paths_list = []
for cl in class_names:
splits = cl.split("_")
subfiles = [f for f in os.listdir(folder + "/" + cl) if ".DS" not in f]
for subf in subfiles:
full_path = os.path.join(folder, cl, subf)
# 229 x 229 is the size resnet 50 uses for images
img = image_utils.load_img(full_path, target_size=(229, 229))
x_raw = image_utils.img_to_array(img)
x_expand = np.expand_dims(
x_raw, axis=0) # add a column for the index of the image
x = preprocess_input(x_expand) # normalize to [-1, 1]
image_list.append(x)
paths_list.append(full_path)
img_data = np.array(image_list)
img_data = np.rollaxis(img_data, 1, 0)
img_data = img_data[0]
return img_data, np.array(labels_list), paths_list
def generate(self):
while True:
#-----------------------------------#
# 对训练集进行打乱
#-----------------------------------#
shuffle_index = np.arange(len(self.imgs_path))
shuffle(shuffle_index)
self.imgs_path = np.array(self.imgs_path, dtype=np.object)[shuffle_index]
self.words = np.array(self.words, dtype=np.object)[shuffle_index]
inputs = []
target0 = []
target1 = []
target2 = []
for i, image_path in enumerate(self.imgs_path):
#-----------------------------------#
# 打开图像，获取对应的标签
#-----------------------------------#
img = Image.open(image_path)
labels = self.words[i]
annotations = np.zeros((0, 15))
for idx, label in enumerate(labels):
annotation = np.zeros((1, 15))
#-----------------------------------#
# bbox 真实框的位置
#-----------------------------------#
annotation[0, 0] = label[0] # x1
annotation[0, 1] = label[1] # y1
annotation[0, 2] = label[0] + label[2] # x2
annotation[0, 3] = label[1] + label[3] # y2
#-----------------------------------#
# landmarks 人脸关键点的位置
#-----------------------------------#
annotation[0, 4] = label[4] # l0_x
annotation[0, 5] = label[5] # l0_y
annotation[0, 6] = label[7] # l1_x
annotation[0, 7] = label[8] # l1_y
annotation[0, 8] = label[10] # l2_x
annotation[0, 9] = label[11] # l2_y
annotation[0, 10] = label[13] # l3_x
annotation[0, 11] = label[14] # l3_y
annotation[0, 12] = label[16] # l4_x
annotation[0, 13] = label[17] # l4_y
if (annotation[0, 4]<0):
annotation[0, 14] = -1
else:
annotation[0, 14] = 1
annotations = np.append(annotations, annotation, axis=0)
target = np.array(annotations)
img, target = get_random_data(img, target, [self.img_size,self.img_size])
# 计算真实框对应的先验框，与这个先验框应当有的预测结果
assignment = self.bbox_util.assign_boxes(target)
regression = assignment[:,:5]
classification = assignment[:,5:8]
landms = assignment[:,8:]
inputs.append(img)
target0.append(np.reshape(regression,[-1,5]))
target1.append(np.reshape(classification,[-1,3]))
target2.append(np.reshape(landms,[-1,10+1]))
if len(target0) == self.batch_size:
tmp_inp = np.array(inputs)
yield preprocess_input(tmp_inp), np.array(target0,dtype=np.float32),np.array(target1,dtype=np.float32),np.array(target2,dtype=np.float32)
inputs = []
target0 = []
target1 = []
target2 = []
def detect_image(self, image_id, image, bgr_img, write_down=False):
# img_ori = image
if write_down:
self.detect_txtfile = open(
"./input/detection-results/" + image_id + ".txt", "w")
image_shape = np.array(np.shape(image)[0:2])
crop_img, x_offset, y_offset = self.letterbox_image(
image, (self.model_image_size[1], self.model_image_size[0]))
photo = np.array(crop_img, dtype=np.float64)
# Image preprocessing
photo = preprocess_input(
np.reshape(
photo,
[1, self.model_image_size[0], self.model_image_size[1], 3]))
start = time.time()
preds = self.ssd_model.predict(photo)
end = time.time()
ti = np.round((end - start) * 1000)
print('Execution time: {} ms'.format(ti))
# Decode the predictions
self.bbox_util = BBoxUtility(
self.num_classes,
overlap_threshold=self.IOU_thresh,
nms_thresh=self.nms_thresh
) # (self, num_classes, priors=None, overlap_threshold=0.5, nms_thresh=0.3, top_k=400)
results = self.bbox_util.detection_out(
preds,
background_label_id=0,
keep_top_k=200,
confidence_threshold=self.confidence)
if len(results[0]) <= 0:
if write_down:
self.detect_txtfile.close()
return bgr_img
det_label = results[0][:, 0]
det_conf = results[0][:, 1]
det_xmin, det_ymin, det_xmax, det_ymax = results[0][:, 2], results[
0][:, 3], results[0][:, 4], results[0][:, 5]
top_indices = [
i for i, conf in enumerate(det_conf) if conf >= self.confidence
]
top_conf = det_conf[top_indices]
top_label_indices = det_label[top_indices].tolist()
top_xmin, top_ymin, top_xmax, top_ymax = np.expand_dims(
det_xmin[top_indices],
-1), np.expand_dims(det_ymin[top_indices], -1), np.expand_dims(
det_xmax[top_indices],
-1), np.expand_dims(det_ymax[top_indices], -1)
# Remove the gray column
boxes = self.ssd_correct_boxes(
top_ymin, top_xmin, top_ymax, top_xmax,
np.array([self.model_image_size[0], self.model_image_size[1]]),
image_shape)
# img_ori = np.array(img_ori)
# img_ori = cv2.cvtColor(bgr_img, cv2.COLOR_RGB2BGR)
img_ori = bgr_img
for i, c in enumerate(top_label_indices):
predicted_class = self.class_names[int(c) - 1]
score = str(top_conf[i])
top, left, bottom, right = boxes[i]
if write_down:
self.detect_txtfile.write(
"%s %s %s %s %s %s\n" %
(predicted_class, score[:6], str(int(left)), str(
int(top)), str(int(right)), str(int(bottom))))
self.plot_one_box(
img_ori,
[int(left), int(top),
int(right), int(bottom)],
label=predicted_class +
', {:.2f}%'.format(np.round(float(score) * 100, 2)),
color=self.colors[int(c) - 1])
cv2.putText(bgr_img,
'FPS: {0} Execution time: {1}ms'.format(
str(int(1000 / ti)), str(int(ti))), (20, 50),
0,
0.5, [20, 150, 255],
thickness=1,
lineType=cv2.LINE_AA)
cv2.imshow('draw', img_ori)
cv2.waitKey(self.gap_time)
if write_down:
self.detect_txtfile.close()
return bgr_img
def preprocess_image(image_path):
img = load_img(image_path, target_size=(224, 224))
img = img_to_array(img)
img = np.expand_dims(img, axis=0)
img = preprocess_input(img)
return img
# load the pre-trained CNN from disk
print("[INFO] loading model...")
model = Model(weights="imagenet")
# load the original image from disk (in OpenCV format) and then
# resize the image to its target dimensions
orig = cv2.imread(args["image"])
resized = cv2.resize(orig, (224, 224))
# load the input image from disk (in Keras/TensorFlow format) and
# preprocess it
image = load_img(args["image"], target_size=(224, 224))
image = img_to_array(image)
image = np.expand_dims(image, axis=0)
image = imagenet_utils.preprocess_input(image)
# use the network to make predictions on the input imag and find
# the class label index with the largest corresponding probability
preds = model.predict(image)
i = np.argmax(preds[0])
# decode the ImageNet predictions to obtain the human-readable label
decoded = imagenet_utils.decode_predictions(preds)
(imagenetID, label, prob) = decoded[0][0]
label = "{}: {:.2f}%".format(label, prob * 100)
print("[INFO] {}".format(label))
# initialize our gradient class activation map and build the heatmap
cam = GradCAM(model, i)
heatmap = cam.compute_heatmap(image)
def predict_quad(model, img, pixel_threshold=cfg.pixel_threshold, quiet=False, img_name=None):
"""
Args:
model: 检测模型，要load_weights的
img: image 图片，文件类型
pixel_threshold: 阈值
quiet:
img_name: 图片的名字
Returns:
text_recs_all：一个列表，每个元素是检测边界的quad值
text_recs_len：text_recs_all的长度，一共检测到多少个区域
img_all: 一个四维数组，img_all[0] 是img_to_array的结果
"""
if not os.path.exists(root_temp):
os.makedirs(root_temp)
if not os.path.exists(root_predict):
os.makedirs(root_predict)
# 获取计算后的图像长宽
d_wight, d_height = resize_image(img, cfg.max_predict_img_size)
# 调整图像大小，便于预测
img = img.resize((d_wight, d_height), Image.BILINEAR).convert('RGB')
img = image.img_to_array(img)
num_img = 1
# 一个4维张量，也就是只有1个3维张量的4维张量
img_all = np.zeros((num_img, d_height, d_wight, 3))
img_all[0] = img
# 将张量的数值大小调到【-1 1】
img_ori = imagenet_utils.preprocess_input(img, mode='tf') # suit tf tensor
# 又整个一样的
x = np.zeros((num_img, d_height, d_wight, 3))
x[0] = img_ori
# (sample, h, w, channels)
y_pred = model.predict(x)
text_recs_all = []
text_recs_len = []
for n in range(num_img):
# (sample, rows, cols, 7_points_pred)
y = y_pred[n]
y[:, :, :3] = sigmoid(y[:, :, :3])
cond = np.greater_equal(y[:, :, 0], pixel_threshold)
activation_pixels = np.where(cond) # fixme 返回元祖tuple类型 a[0]保存了纵坐标 a[1]保存横坐标
quad_scores, quad_after_nms = nms(y, activation_pixels)
text_recs = []
x[n] = np.uint8(x[n])
with image.array_to_img(img_all[n]) as im: # Image.fromarray(x[n]) error ?
im_array = x[n]
# fixme 注意：拿去CRNN识别的是缩放后的图像
scale_ratio_w = 1
scale_ratio_h = 1
quad_im = im.copy()
draw = ImageDraw.Draw(im)
# 拷贝一个原图像，在拷贝的图像上有文字的地方画线
for i, j in zip(activation_pixels[0], activation_pixels[1]):
px = (j + 0.5) * cfg.pixel_size
py = (i + 0.5) * cfg.pixel_size
line_width, line_color = 1, 'blue'
if y[i, j, 1] >= cfg.side_vertex_pixel_threshold:
if y[i, j, 2] < cfg.trunc_threshold:
line_width, line_color = 2, 'yellow'
elif y[i, j, 2] >= 1 - cfg.trunc_threshold:
line_width, line_color = 2, 'green'
draw.line([(px - 0.5 * cfg.pixel_size, py - 0.5 * cfg.pixel_size),
(px + 0.5 * cfg.pixel_size, py - 0.5 * cfg.pixel_size),
(px + 0.5 * cfg.pixel_size, py + 0.5 * cfg.pixel_size),
(px - 0.5 * cfg.pixel_size, py + 0.5 * cfg.pixel_size),
(px - 0.5 * cfg.pixel_size, py - 0.5 * cfg.pixel_size)],
width=line_width, fill=line_color)
if not img_name is None:
im.save(root_temp + img_name + '.jpg')
quad_draw = ImageDraw.Draw(quad_im)
for score, geo, s in zip(quad_scores, quad_after_nms,
range(len(quad_scores))):
if np.amin(score) > 0:
quad_draw.line([tuple(geo[0]),
tuple(geo[1]),
tuple(geo[2]),
tuple(geo[3]),
tuple(geo[0])], width=2, fill='blue')
if cfg.predict_cut_text_line:
cut_text_line(geo, scale_ratio_w, scale_ratio_h, im_array,
img_name, s)
rescaled_geo = geo / [scale_ratio_w, scale_ratio_h]
text_rec = np.reshape(rescaled_geo, (8,)).tolist()
text_recs.append(text_rec)
elif not quiet:
print('quad invalid with vertex num less then 4.')
if not img_name is None:
quad_im.save(root_predict + img_name + '.jpg' )
for t in range(len(text_recs)):
text_recs_all.append(text_recs[t])
text_recs_len.append(len(text_recs))
return text_recs_all, text_recs_len, img_all
def predict(east_detect, img_path, pixel_threshold, quiet=False):
img = image.load_img(img_path)
d_wight, d_height = resize_image(img, cfg.max_predict_img_size)
img = img.resize((d_wight, d_height), Image.NEAREST).convert('RGB')
img = image.img_to_array(img)
# 将张量的值 调到【-1 1】
img = imagenet_utils.preprocess_input(img,mode='tf')
# 变成4维张量
x = np.expand_dims(img, axis=0)
y = east_detect.predict(x)
y = np.squeeze(y, axis=0)
y[:, :, :3] = sigmoid(y[:, :, :3])
cond = np.greater_equal(y[:, :, 0], pixel_threshold)
activation_pixels = np.where(cond)
quad_scores, quad_after_nms = nms(y, activation_pixels)
with Image.open(img_path) as im:
im_array = image.img_to_array(im.convert('RGB'))
d_wight, d_height = resize_image(im, cfg.max_predict_img_size)
scale_ratio_w = d_wight / im.width
scale_ratio_h = d_height / im.height
im = im.resize((d_wight, d_height), Image.NEAREST).convert('RGB')
quad_im = im.copy()
draw = ImageDraw.Draw(im)
for i, j in zip(activation_pixels[0], activation_pixels[1]):
px = (j + 0.5) * cfg.pixel_size
py = (i + 0.5) * cfg.pixel_size
line_width, line_color = 1, 'red'
if y[i, j, 1] >= cfg.side_vertex_pixel_threshold:
if y[i, j, 2] < cfg.trunc_threshold:
line_width, line_color = 2, 'yellow'
elif y[i, j, 2] >= 1 - cfg.trunc_threshold:
line_width, line_color = 2, 'green'
draw.line([(px - 0.5 * cfg.pixel_size, py - 0.5 * cfg.pixel_size),
(px + 0.5 * cfg.pixel_size, py - 0.5 * cfg.pixel_size),
(px + 0.5 * cfg.pixel_size, py + 0.5 * cfg.pixel_size),
(px - 0.5 * cfg.pixel_size, py + 0.5 * cfg.pixel_size),
(px - 0.5 * cfg.pixel_size, py - 0.5 * cfg.pixel_size)],
width=line_width, fill=line_color)
im.save(img_path + '_act.jpg')
quad_draw = ImageDraw.Draw(quad_im)
txt_items = []
for score, geo, s in zip(quad_scores, quad_after_nms,
range(len(quad_scores))):
if np.amin(score) > 0:
quad_draw.line([tuple(geo[0]),
tuple(geo[1]),
tuple(geo[2]),
tuple(geo[3]),
tuple(geo[0])], width=2, fill='red')
if cfg.predict_cut_text_line:
cut_text_line(geo, scale_ratio_w, scale_ratio_h, im_array,
img_path, s)
rescaled_geo = geo / [scale_ratio_w, scale_ratio_h]
rescaled_geo_list = np.reshape(rescaled_geo, (8,)).tolist()
txt_item = ','.join(map(str, rescaled_geo_list))
txt_items.append(txt_item + '\n')
elif not quiet:
print('quad invalid with vertex num less then 4.')
quad_im.save(img_path + '_predict.jpg')
if cfg.predict_write2txt and len(txt_items) > 0:
with open(img_path[:-4] + '.txt', 'w') as f_txt:
f_txt.writelines(txt_items)
def detect_image(self, image):
old_image = image.copy()
image = np.array(image, np.float32)
im_height, im_width, _ = np.shape(image)
scale = [im_width, im_height, im_width, im_height]
scale_for_landmarks = [
im_width, im_height, im_width, im_height, im_width, im_height,
im_width, im_height, im_width, im_height
]
# 图片预处理，归一化
photo = np.expand_dims(preprocess_input(image), 0)
anchors = Anchors(self.cfg,
image_size=(im_height, im_width)).get_anchors()
preds = self.get_pred(photo)
preds = [pred.numpy() for pred in preds]
# 将预测结果进行解码和非极大抑制
results = self.bbox_util.detection_out(
preds, anchors, confidence_threshold=self.confidence)
if len(results) <= 0:
return old_image, 0, 0
results = np.array(results)
results[:, :4] = results[:, :4] * scale
results[:, 5:] = results[:, 5:] * scale_for_landmarks
for b in results:
text = "{:.4f}".format(b[4])
b = list(map(int, b))
cv2.rectangle(old_image, (b[0], b[1]), (b[2], b[3]), (0, 0, 255),
2)
#####################
global cnt, t0, t1
t1 = time.time()
image_clip = old_image
# if t1 - t0 > 1:
# t0 = t1
# image_clip = old_image[b[1]-20:b[3]+20, b[0]-20:b[2]+20]
image_clip = old_image[b[1]:b[3], b[0]:b[2]]
image_clip = cv2.cvtColor(image_clip, cv2.COLOR_RGB2BGR)
# 保存剪切的图片
# cv2.imshow("clip", image_clip)
# cv2.imwrite("savedImg/wang/" + str(t1) + ".png", image_clip)
# cnt += 1
# print(cnt)
#####################
cx = b[0]
cy = b[1] + 12
cv2.putText(old_image, text, (cx, cy), cv2.FONT_HERSHEY_DUPLEX,
0.5, (255, 255, 255))
# landms
cv2.circle(old_image, (b[5], b[6]), 1, (0, 0, 255), 4)
cv2.circle(old_image, (b[7], b[8]), 1, (0, 255, 255), 4)
cv2.circle(old_image, (b[9], b[10]), 1, (255, 0, 255), 4)
cv2.circle(old_image, (b[11], b[12]), 1, (0, 255, 0), 4)
cv2.circle(old_image, (b[13], b[14]), 1, (255, 0, 0), 4)
return old_image, image_clip, len(results)
print(folder, z)
pathn = path + folder
os.chdir(pathn)
filenames = os.listdir(pathn)
for f in filenames:
x = cv2.imread(f)
x = cv2.resize(x, (224, 224), interpolation=cv2.INTER_CUBIC)
if z == 1 or z == 2:
D = d_aug(x)
else:
D = [x]
# D=D[:3]
for x in D:
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
x = x / 255
img_data.append(x)
if z == 2:
labels.append(1)
else:
labels.append(z)
img_data = np.array(img_data)
img_data = img_data.astype('float32')
print(img_data.shape)
img_data = np.rollaxis(img_data, 1, 0)
print(img_data.shape)
img_data = img_data[0]
print(img_data.shape)
# labels = np.expand_dims(labels, axis=0)
pil_input_images = [] # Store resized versions of the images here.
# We'll only load one image in this example.
img_dir = './example_images'
write_out_dir = './results'
os.makedirs(write_out_dir, exist_ok=True)
for image_path in glob.glob(os.path.join(img_dir, '*.jpg')):
pil_image = image.load_img(image_path,
target_size=(config.height, config.width))
pil_input_images.append(pil_image)
img = np.array(pil_image)
input_images.append(img)
input_images = preprocess_input(np.array(input_images))
bboxes, scores, labels = prediction_model.predict(input_images)
confidence_threshold = 0.75
for index in range(input_images.shape[0]):
bbox = bboxes[index]
confidence = scores[index]
label = labels[index]
print(bbox.shape)
# confidence
annotated_image = image_annotator(image=pil_input_images[index],
def upload_file():
#print('here')
if request.method == 'POST':
# check if the post request has the file part
if 'file' not in request.files:
flash('No file part')
return redirect(request.url)
file = request.files['file']
if file.filename == '':
flash('No file selected for uploading')
return redirect(request.url)
if file and allowed_file(file.filename):
global filename
filename = secure_filename(file.filename)
img_path = os.path.join(app.config['UPLOAD_FOLDER'], filename)
file.save(img_path)
# Get Features of uploaded image
# check if we alrey have this file (by name) in our database
have_file = database_df[database_df.loc[:, 'Filename'] == filename]
if (have_file.shape[0] >
0): # we have the features in the database
img_features = have_file.image_features.tolist()
else:
img_data = image.load_img(img_path,
target_size=(pwidth, pheight))
img_vector = image.img_to_array(img_data)
img_vector = np.expand_dims(img_vector, axis=0)
img_vector = preprocess_input(
img_vector
) #Problem here, must be convention of keras to pass by reference?
img_vector = imagenet_utils.preprocess_input(img_vector)
img_features = network_model.predict(img_vector)
#print(type(img_features))
# get 9 nearest neighbors
# n_neighs = 9
# nn_index, neighbors,distance = nearest_neighbor_image_finder(img_features, n_neighs, database_df)
dist, nn_index = neighs.kneighbors(img_features,
return_distance=True)
distance = dist.tolist()[0]
# fix path to the database...
neighbors = database_df.iloc[nn_index.tolist()[0]].copy()
neighbors.loc[:, 'db_path'] = neighbors.loc[:, 'path'].astype(
str).copy()
#neighbors = neighbors_db # this line is where the "filtering" should occur if we add handles on website
#image_name = os.path.join('images',filename)
# print(f'saved: {img_path}')
# print(f'<upload path>: {UPLOAD_FOLDER}')
# print(f'image: {filename}')
npath = neighbors['db_path'][0]
# print(f'saved: {npath}')
# print(f'<media path>: {MEDIA_FOLDER}')
header_copy = 'your example:'
return render_template('album.html',
header_copy=header_copy,
image_name=filename,
neighbors=neighbors,
dist=distance)
vgg_face = Model(inputs=model.layers[0].input, outputs=model.layers[-2].output)
x_train = []
y_train = []
person_rep = dict()
person_folders = os.listdir('Images_crop')
for i, person in enumerate(person_folders):
person_rep[i] = person
image_names = os.listdir('Images_crop/' + person + '/')
for image_name in image_names:
img = load_img('Images_crop/' + person + '/' + image_name,
target_size=(224, 224))
img = img_to_array(img)
img = np.expand_dims(img, axis=0)
img = preprocess_input(img)
img_encode = vgg_face(img)
x_train.append(np.squeeze(K.eval(img_encode)).tolist())
y_train.append(i)
# Prepare Test Data
x_test = []
y_test = []
test_image_names = os.listdir('Test_Images_crop/' + person + '/')
for image_name in test_image_names:
img = load_img('Test_Images_crop/' + person + '/' + image_name,
target_size=(224, 224))
img = img_to_array(img)
img = np.expand_dims(img, axis=0)
img = preprocess_input(img)
img_encode = vgg_face(img)
def get_FPS(self, image, test_interval):
#-------------------------------------#
# 转换成RGB图片，可以用于灰度图预测。
#-------------------------------------#
image = image.convert("RGB")
image_shape = np.array(np.shape(image)[0:2])
old_width, old_height = image_shape[1], image_shape[0]
#---------------------------------------------------------#
# 给原图像进行resize，resize到短边为600的大小上
#---------------------------------------------------------#
width, height = get_new_img_size(old_width, old_height)
image = image.resize([width,height], Image.BICUBIC)
photo = np.array(image,dtype = np.float64)
#-----------------------------------------------------------#
# 图片预处理，归一化。
#-----------------------------------------------------------#
photo = preprocess_input(np.expand_dims(photo,0))
rpn_pred = self.model_rpn.predict(photo)
#-----------------------------------------------------------#
# 将建议框网络的预测结果进行解码
#-----------------------------------------------------------#
base_feature_width, base_feature_height = self.get_img_output_length(width, height)
anchors = get_anchors([base_feature_width, base_feature_height], width, height)
rpn_results = self.bbox_util.detection_out_rpn(rpn_pred, anchors)
#-------------------------------------------------------------#
# 在获得建议框和共享特征层后，将二者传入classifier中进行预测
#-------------------------------------------------------------#
base_layer = rpn_pred[2]
proposal_box = np.array(rpn_results)[:, :, 1:]
temp_ROIs = np.zeros_like(proposal_box)
temp_ROIs[:, :, [0, 1, 2, 3]] = proposal_box[:, :, [1, 0, 3, 2]]
classifier_pred = self.model_classifier.predict([base_layer, temp_ROIs])
#-------------------------------------------------------------#
# 利用classifier的预测结果对建议框进行解码，获得预测框
#-------------------------------------------------------------#
results = self.bbox_util.detection_out_classifier(classifier_pred, proposal_box, self.config, self.confidence)
if len(results[0])>0:
results = np.array(results[0])
boxes = results[:, :4]
top_conf = results[:, 4]
top_label_indices = results[:, 5]
boxes[:, [0, 2]] = boxes[:, [0, 2]] * old_width
boxes[:, [1, 3]] = boxes[:, [1, 3]] * old_height
t1 = time.time()
for _ in range(test_interval):
rpn_pred = self.model_rpn.predict(photo)
#-----------------------------------------------------------#
# 将建议框网络的预测结果进行解码
#-----------------------------------------------------------#
base_feature_width, base_feature_height = self.get_img_output_length(width, height)
anchors = get_anchors([base_feature_width, base_feature_height], width, height)
rpn_results = self.bbox_util.detection_out_rpn(rpn_pred, anchors)
#-------------------------------------------------------------#
# 在获得建议框和共享特征层后，将二者传入classifier中进行预测
#-------------------------------------------------------------#
base_layer = rpn_pred[2]
proposal_box = np.array(rpn_results)[:, :, 1:]
temp_ROIs = np.zeros_like(proposal_box)
temp_ROIs[:, :, [0, 1, 2, 3]] = proposal_box[:, :, [1, 0, 3, 2]]
classifier_pred = self.model_classifier.predict([base_layer, temp_ROIs])
#-------------------------------------------------------------#
# 利用classifier的预测结果对建议框进行解码，获得预测框
#-------------------------------------------------------------#
results = self.bbox_util.detection_out_classifier(classifier_pred, proposal_box, self.config, self.confidence)
if len(results[0])>0:
results = np.array(results[0])
boxes = results[:, :4]
top_conf = results[:, 4]
top_label_indices = results[:, 5]
boxes[:, [0, 2]] = boxes[:, [0, 2]] * old_width
boxes[:, [1, 3]] = boxes[:, [1, 3]] * old_height
t2 = time.time()
tact_time = (t2 - t1) / test_interval
return tact_time
def prepare_input(file_path, keypoints, image_size=IMAGE_SIZE):
img = tf.io.read_file(file_path)
img = decode_img(img, image_size)
img = preprocess_input(img, mode='torch')
group_keypoints = tf.cast(tf.reshape(keypoints, shape=(n_keypoints, 2)), tf.float32)
return img, group_keypoints