Skip to content

2
2020-077
Commit d15b632b authored by Manoj Kumar's avatar Manoj Kumar
Browse files
Options

code cleanup

parent a9558498
Hide whitespace changes
Inline Side-by-side
Showing with 39 additions and 463 deletions
dataq/detect.py
View file @ d15b632b
# * The detect.py file is created to implement the front-end of
# * English to SSL Translator
# *
# * @author Manoj Kumar | IT17050272
# * @version 1.0
# * @since 2020-09-12
#imports
import numpy as np
import os
import sys
......@@ -8,26 +16,23 @@ from PIL import Image
from tensorflow import keras
sys.path.append('.')
from translation.modelUp import sign_predict
from object_detection.utils import label_map_util
from object_detection.utils import visualization_utils as vis_util
#Set Base Path
PATH = os.path.dirname(os.path.abspath(__file__))
#Check TensorFlow version
if StrictVersion(tf.__version__) < StrictVersion('1.15.3'):
raise ImportError('Please upgrade your TensorFlow installation to v1.15.3 or later!')
print(tf.__version__)
from object_detection.utils import label_map_util
from object_detection.utils import visualization_utils as vis_util
# Set Model Paths
MODEL_NAME = PATH +'\inference_graph'
PATH_TO_FROZEN_GRAPH = MODEL_NAME + '\\frozen_inference_graph.pb'
PATH_TO_LABELS = os.path.join(PATH +'\\training','labelmap.pbtxt')
#load tensorflow model to memory
detection_graph = tf.Graph()
with detection_graph.as_default():
......@@ -41,6 +46,7 @@ with detection_graph.as_default():
#map labels
category_index = label_map_util.create_category_index_from_labelmap(PATH_TO_LABELS, use_display_name=True)
#run inference for a single frame
def run_inference_for_single_image(image, graph):
if 'detection_masks' in tensor_dict:
# The following processing is only for single image
......@@ -55,8 +61,7 @@ def run_inference_for_single_image(image, graph):
detection_masks_reframed = tf.cast(tf.greater(detection_masks_reframed, 0.5), tf.uint8)
# Follow the convention by adding back the batch dimension
tensor_dict['detection_masks'] = tf.expand_dims(
detection_masks_reframed, 0)
tensor_dict['detection_masks'] = tf.expand_dims(detection_masks_reframed, 0)
image_tensor = tf.compat.v1.get_default_graph().get_tensor_by_name('image_tensor:0')
# Run inference
......@@ -65,16 +70,16 @@ def run_inference_for_single_image(image, graph):
# outputs are float32 numpy arrays
output_dict['num_detections'] = int(output_dict['num_detections'][0])
output_dict['detection_classes'] = output_dict[
'detection_classes'][0].astype(np.uint8)
output_dict['detection_classes'] = output_dict['detection_classes'][0].astype(np.uint8)
output_dict['detection_boxes'] = output_dict['detection_boxes'][0]
output_dict['detection_scores'] = output_dict['detection_scores'][0]
if 'detection_masks' in output_dict:
output_dict['detection_masks'] = output_dict['detection_masks'][0]
return output_dict
#initialize camera
cap = cv2.VideoCapture(0)
try:
with detection_graph.as_default():
with tf.compat.v1.Session() as sess:
......@@ -112,14 +117,16 @@ try:
line_thickness=8)
score = round(100*output_dict['detection_scores'][0])
#send the request to bavan here
#send the request to translation component here
###
# I will be sending a POST request to u. a hand picture
if score > 80:
print(image_np_expanded.shape)
# sign_predict(image_np)
# sign_predict(image_np) <-- still under development
#waiting for the API on that component to be built
# end send request
## Press Q to close the camera
cv2.imshow('Hand Detector. Press Q to close', cv2.resize(image_np, (800, 600)))
if cv2.waitKey(25) & 0xFF == ord('q'):
......
dataq/prepare.py deleted 100644 → 0
View file @ a9558498
import os,cv2,keras
import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
import tensorflow as tf
from skimage import io
import json
path = "dataq\\images"
annot = "Airplanes_Annotations"
#load CSV
train = pd.read_csv('./dataq/train.csv',encoding='utf-8')
print('CSV Loaded')
print(tf.__version__)
#put bounding box
for index,row in train.iterrows():
filename = row["External ID"]
print('File Name:' + filename)
image = io.imread(train['Labeled Data'][index])
labels = row['Label']
label = json.loads(labels)
label_obj_json = label['objects'][0]
bbox = label_obj_json['bbox']
height = bbox['height']
width = bbox['width']
xmax = bbox['left'] + width
ymax = bbox['top'] + height
cv2.rectangle(image, (bbox['left'], bbox['top']), (xmax, ymax), (255, 0, 0), 2)
plt.figure()
plt.imshow(image)
plt.show()
break
cv2.setUseOptimized(True);
ss = cv2.ximgproc.segmentation.createSelectiveSearchSegmentation()
im = cv2.imread(os.path.join(path,"image_001.jpg"))
ss.setBaseImage(im)
ss.switchToSelectiveSearchFast()
rects = ss.process()
imOut = im.copy()
for i, rect in (enumerate(rects)):
x, y, w, h = rect
# print(x,y,w,h)
# imOut = imOut[x:x+w,y:y+h]
cv2.rectangle(imOut, (x, y), (x+w, y+h), (0, 255, 0), 1, cv2.LINE_AA)
# plt.figure()
plt.imshow(imOut)
plt.show()
train_images=[]
train_labels=[]
# get intersection over union
def get_iou(bb1, bb2):
assert bb1['x1'] < bb1['x2']
assert bb1['y1'] < bb1['y2']
assert bb2['x1'] < bb2['x2']
assert bb2['y1'] < bb2['y2']
print('inside get IOU')
x_left = max(bb1['x1'], bb2['x1'])
y_top = max(bb1['y1'], bb2['y1'])
x_right = min(bb1['x2'], bb2['x2'])
y_bottom = min(bb1['y2'], bb2['y2'])
if x_right < x_left or y_bottom < y_top:
return 0.0
intersection_area = (x_right - x_left) * (y_bottom - y_top)
bb1_area = (bb1['x2'] - bb1['x1']) * (bb1['y2'] - bb1['y1'])
bb2_area = (bb2['x2'] - bb2['x1']) * (bb2['y2'] - bb2['y1'])
iou = intersection_area / float(bb1_area + bb2_area - intersection_area)
assert iou >= 0.0
assert iou <= 1.0
return iou
#again? IDK Why
ss = cv2.ximgproc.segmentation.createSelectiveSearchSegmentation()
listOfImages = []
for imageFiles in os.listdir(path):
listOfImages.append(imageFiles)
print(listOfImages)
#add bounding boxes
for index,row in train.iterrows():
image_name = row["External ID"]
try:
if image_name.startswith("image") and image_name in listOfImages:
gtvalues = []
print("Image Name:"+ image_name)
image = cv2.imread((os.path.join(path,image_name)))
labels = row['Label']
label = json.loads(labels)
label_obj_json = label['objects'][0]
bbox = label_obj_json['bbox']
height = bbox['height']
width = bbox['width']
x1 = bbox['left']
y1 = bbox['top']
x2 = x1 + width
y2 = x2 + height
gtvalues.append({
"x1": x1,
"x2":x2,
"y1":y1,
"y2":y2
})
ss.setBaseImage(image)
ss.setBaseImage(image)
ss.switchToSelectiveSearchFast()
ssresults = ss.process()
imout = image.copy()
counter = 0
falsecounter = 0
flag = 0
fflag = 0
bflag = 0
print(ssresults)
for e, result in enumerate(ssresults):
if e < 2000 and flag == 0:
for gtval in gtvalues:
x,y,w,h = result
iou = get_iou(gtval,{"x1":x1,"x2":x2,"y1":y1,"y2":y2})
if counter < 30:
if iou < 0.70:
timage = imOut[y:y+h, x: x+w]
resized = cv2.resize(timage,(224,224), interpolation=cv2.INTER_AREA)
train_images.append(1)
counter+=1
else:
fflag = 1
if falsecounter < 30:
if iou < 0.30:
timage = imout[y:y+h,x:x+w]
resized = cv2.resize(timage, (224,224), interpolation = cv2.INTER_AREA)
train_images.append(resized)
train_labels.append(0)
falsecounter += 1
else:
bflag = 1
if fflag == 1 and bflag == 1:
print("inside")
flag = 1
#print(image_name)
except Exception as e:
print(e)
continue
X_new = np.array(train_images)
y_new = np.array(train_labels)
print('Shape of X_new:')
X_new.shape
from keras.layers import Dense
from keras import Model
from keras import optimizers
from keras.preprocessing.image import ImageDataGenerator
from keras.applications.vgg16 import VGG16
vggmodel = VGG16(weights='imagenet', include_top=True)
vggmodel.summary()
for layers in (vggmodel.layers)[:15]:
print(layers)
layers.trainable = False
X = vggmodel.layers[-2].output
predictions = Dense(2, activation="softmax")(X)
model_final = Model(inputs = vggmodel.input, outputs = predictions)
from keras.optimizers import Adam
opt = Adam(learning_rate=0.0001)
model_final.compile(loss= keras.losses.categorical_crossentropy, optimizer= opt, metrics=["accuracy"])
model_final.summary()
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelBinarizer
class MyLabelBinarizer(LabelBinarizer):
def transform(self, y):
Y = super().transform(y)
if self.y_type_ == 'binary':
return np.hstack((Y,1-Y))
else:
return Y
def inverse_transform(self, Y, threshold=None):
if self.y_type_ == 'binary':
return super().inverse_transform(Y[:, 0], threshold)
else:
return super().inverse_transform(Y, threshold=threshold)
lenc = MyLabelBinarizer()
Y = lenc.fit_transform(y_new)
X_train, X_test, y_train, y_test = train_test_split(X_new,Y,test_size=0.25)
print(X_train.shape,X_test.shape,y_train.shape,y_test.shape)
trdata = ImageDataGenerator(horizontal_flip=True, vertical_flip=True, rotation_range=90)
traindata = trdata.flow(x=X_train, y=y_train)
tsdata = ImageDataGenerator(horizontal_flip=True, vertical_flip=True, rotation_range=90)
testdata = tsdata.flow(x=X_test, y=y_test)
from keras.callbacks import ModelCheckpoint, EarlyStopping
checkpoint = ModelCheckpoint("ieeercnn_vgg16_1.h5", monitor='val_loss', verbose=1, save_best_only=True, save_weights_only=False, mode='auto', period=1)
early = EarlyStopping(monitor='val_loss', min_delta=0, patience=100, verbose=1, mode='auto')
hist = model_final.fit_generator(generator= traindata, steps_per_epoch= 10, epochs= 1000, validation_data= testdata, validation_steps=2, callbacks=[checkpoint,early])
import matplotlib.pyplot as plt
# plt.plot(hist.history["acc"])
# plt.plot(hist.history['val_acc'])
plt.plot(hist.history['loss'])
plt.plot(hist.history['val_loss'])
plt.title("model loss")
plt.ylabel("Loss")
plt.xlabel("Epoch")
plt.legend(["Loss","Validation Loss"])
plt.show()
plt.savefig('chart loss.png')
im = X_test[1600]
plt.imshow(im)
img = np.expand_dims(im, axis=0)
out= model_final.predict(img)
if out[0][0] > out[0][1]:
print("hand")
else:
print("not a hand")
#### =====================TESTING=================#####
# z=0
# for e,i in enumerate(os.listdir(path)):
# if i.startswith("4"):
# z += 1
# img = cv2.imread(os.path.join(path,i))
# ss.setBaseImage(img)
# ss.switchToSelectiveSearchFast()
# ssresults = ss.process()
# imout = img.copy()
# for e,result in enumerate(ssresults):
# if e < 2000:
# x,y,w,h = result
# timage = imout[y:y+h,x:x+w]
# resized = cv2.resize(timage, (224,224), interpolation = cv2.INTER_AREA)
# img = np.expand_dims(resized, axis=0)
# out= model_final.predict(img)
# if out[0][0] > 0.65:
# cv2.rectangle(imout, (x, y), (x+w, y+h), (0, 255, 0), 1, cv2.LINE_AA)
# plt.figure()
# plt.imshow(imout)
\ No newline at end of file
dataq/test.py deleted 100644 → 0
View file @ a9558498
import os
import tensorflow as tf
from tensorflow import keras
import numpy as np
import matplotlib.pyplot as plt
import cv2
import pandas as pd
from sklearn.model_selection import train_test_split
from sklearn.metrics import confusion_matrix
print("Tensorflow Version: " + tf.__version__)
imagepaths = []
for root,dirs,files in os.walk(".",topdown=False):
for name in files:
path = os.path.join(root,name)
if path.endswith("jpg"):
imagepaths.append(path)
print(len(imagepaths))
def plot_image(path):
img = cv2.imread(path)
img_cvt = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
print(img_cvt.shape)
plt.grid(False)
plt.imshow(img_cvt)
plt.xlabel("Width")
plt.ylabel("Height")
plt.title("Image" + path)
plt.show()
plot_image(imagepaths[0])
X = []
y = []
for path in imagepaths:
image = cv2.imread(path)
image = cv2.cvtColor(image,cv2.COLOR_BGR2GRAY)
image = cv2.resize(image,(320,120))
X.append(image)
label = int(path.split("\\")[3])
y.append(label)
X = np.array(X,dtype="uint8")
X = X.reshape(len(imagepaths),120,320,1)
y = np.array(y)
print("Images Loaded: ",len(X))
print("Labels Loaded: ", len(y))
print(y[0], imagepaths[0])
#TRAINING AND TESTING
test_score = 0.2
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=test_score, random_state=100)
print(y_test)
from keras.models import Sequential
from keras.layers.convolutional import Conv2D, MaxPooling2D
from keras.layers import Dense, Flatten
model = Sequential()
model.add(Conv2D(32, (5, 5), activation='relu', input_shape=(120, 320, 1)))
model.add(MaxPooling2D((2, 2)))
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(MaxPooling2D((2, 2)))
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(MaxPooling2D((2, 2)))
model.add(Flatten())
model.add(Dense(128, activation='relu'))
model.add(Dense(10, activation='softmax'))
model.compile(optimizer='adam',loss='sparse_categorical_crossentropy',metrics=['accuracy'])
model.fit(X_train, y_train, epochs=5, batch_size=64, verbose=2, validation_data=(X_test, y_test))
model.save('./dataq/handrecognition_model.h5')
# model = keras.models.load_model('handrecognition_model.h5')
# model.summary()
#testing
test_loss, test_acc = model.evaluate(X_test, y_test)
print('Test accuracy: {:2.2f}%'.format(test_acc*100))
cam = cv2.VideoCapture(0)
# while True:
# ret, image = cam.read()
# img = tf.keras.preprocessing.image.img_to_array(image)
# predictions = model.predict(img) # Make predictions towards the test set
# cv2.imshow('frame',frame)
# if cv2.waitKey(1) & 0xFF == ord('q'):
# break
predictions = model.predict(X_test) # Make predictions towards the test set
print(predictions)
print(np.argmax(predictions[0]), y_test[0])
def validate_9_images(predictions_array, true_label_array, img_array):
# Array for pretty printing and then figure size
class_names = ["no-hand", "hand"]
plt.figure(figsize=(15,5))
for i in range(1, 10):
# Just assigning variables
prediction = predictions_array[i]
true_label = true_label_array[i]
img = img_array[i]
img = cv2.cvtColor(img, cv2.COLOR_GRAY2RGB)
# Plot in a good way
plt.subplot(3,3,i)
plt.grid(False)
plt.xticks([])
plt.yticks([])
plt.imshow(img, cmap=plt.cm.binary)
predicted_label = np.argmax(prediction) # Get index of the predicted label from prediction
# Change color of title based on good prediction or not
if predicted_label == true_label:
color = 'blue'
else:
color = 'red'
plt.xlabel("Predicted: {} {:2.0f}% (True: {})".format(class_names[predicted_label],
100*np.max(prediction),
class_names[true_label]),
color=color)
plt.show()
validate_9_images(predictions, y_test, X_test)
y_pred = np.argmax(predictions, axis=1) # Transform predictions into 1-D array with label number
print(y_pred)
df = pd.DataFrame(confusion_matrix(y_test, y_pred),
columns=["Predicted No Hand", "Predicted Hand"],
index=["Actual No Hand", "Actual Hand"])
print(df)
print("Done!")
\ No newline at end of file
dataq/xml_to_csv.py
View file @ d15b632b
# * The detect.py file is created to implement the front-end of
# * English to SSL Translator
# *
# * @author Manoj Kumar | IT17050272
# * @version 1.0
# * @since 2020-09-23
#imports
import os
import glob
import pandas as pd
import xml.etree.ElementTree as ET
#converter *.xml file to *.csv
def xml_to_csv(path):
xml_list = []
#Check all XML files in current directory
for xml_file in glob.glob(path + '/*.xml'):
tree = ET.parse(xml_file)
root = tree.getroot()
......@@ -20,17 +30,21 @@ def xml_to_csv(path):
int(member[4][3].text)
)
xml_list.append(value)
#predefined column name
column_name = ['filename', 'width', 'height', 'class', 'xmin', 'ymin', 'xmax', 'ymax']
xml_df = pd.DataFrame(xml_list, columns=column_name)
return xml_df
def converter():
#check directories : train and test
for directory in ['train','test']:
image_path = os.path.join(os.getcwd(), 'images/{}'.format(directory))
xml_df = xml_to_csv(image_path)
xml_df.to_csv('images/{}_labels.csv'.format(directory), index=None)
print('Successfully converted xml to csv.')
print("xmltocsv")
#main method
converter()
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment