Added oneshot import package

Author: Thanatoz-1

oneshot.py

@@ -0,0 +1,141 @@
+import os
+import keras
+import cv2
+from urllib.request import urlretrieve
+from tqdm import tqdm
+import pandas as pd 
+import numpy as np
+import json
+
+def triplet_loss(y_true, y_pred, alpha = 0.2):
+	'''
+	Returns the triplet loss.
+	'''
+    anchor, positive, negative = y_pred[0], y_pred[1], y_pred[2]
+    pos_dist = tf.reduce_sum(tf.square(anchor - positive))
+    neg_dist = tf.reduce_sum(tf.square(anchor - negative))
+    basic_loss = pos_dist - neg_dist + alpha
+    loss = tf.maximum(basic_loss, 0)    
+    return loss
+
+
+def init_model(input_shape=(200,200,3)):
+	'''
+	Initalizes the model and returns the model
+
+	Arguments:
+	input_shape: shape of image input along with channels
+	Default is (200,200,3) and if you wish to change, it should be greater than this.
+	'''
+	assert(input_shape[0]<200 or input_shape[1]<200), 'Shape cannot be less than 200'
+	assert(input_shape[2]!=3), 'RGB channels required'
+	RecModel = InceptionV3(input_shape=input_shape, weights=None, include_top=False)
+	RecModel.compile(optimizer='adam', loss = triplet_loss, metrics = ['accuracy'])
+	return RecModel
+
+#Image encoding. The main aim of this function is to convert the images to encodings.
+def image_to_encoding(image_name, model):
+	'''
+	Images to encoding runs one forward propogation on the image and creates encodings.
+
+	Arguments:
+	image_name: Image name is the name of image to encode with extension
+	model: The model on which we have to forward propogate
+	'''
+
+    #Read the image and convert BGR to RGB
+    a = cv2.imread('./images/'+str(image_name),cv2.COLOR_BGR2RGB)
+    #Generate the image padding using np.zeros
+    b = np.zeros((200,200,3), dtype='uint8')
+    #Add the slice of image over the padding
+    b[:a.shape[0],:a.shape[1],:]+=a
+    #Convert to batch
+    b = b.reshape(1,200,200,3)
+    #Forward propagate through the model network
+    embedding = model.predict_on_batch(b)
+    return embedding
+
+def create_encodings(path_to_images='./images'):
+	'''
+	This generates embeddings and take 0.15s for each element of dataset
+	So use accordingly.
+
+	Arguments:
+	path_to_image = pass the path to image string to the function.
+	'''
+	imgs = os.listdir('./images')
+	database={}
+	for i,img in enumerate(sorted(imgs)):
+	    database[img.split('.')[0]]=image_to_encoding(img, RecModel)
+    return database
+
+#Function to verify the images with the identity
+def verify(image,identity, database, model):
+    """
+    Function that verifies Redundant images
+    
+    Arguments:
+    image_path -- path to an image
+    identity -- string, productId you'd like to verify the identity.
+    database -- python dictionary mapping Products to their encodings (vectors).
+    model -- your ResNet model instance in Keras
+    
+    Returns:
+    dist -- distance between the image_path and the image of "identity" in the database.
+    same -- True, if the images are same. False otherwise.
+    """
+    
+    # Step 1: Compute the encoding for the image. Use image_to_encoding()
+    encoding = image_to_encoding(image, model)
+    # Step 2: Compute distance with identity's image (≈ 1 line)
+    dist = np.linalg.norm(database[identity] - encoding)
+    
+    # Step 3: Open the door if dist < 0.7, else don't open (≈ 3 lines)
+    if dist < 1:
+        same = True
+    else:
+        same = False
+    return dist, same
+
+if __name__ == '__main__':
+	if not os.path.exists('./dataset'):
+		os.mkdir('./dataset')
+	if not os.path.exists('./images/'):
+		try:
+			link_zip='https://transfer.sh/ShBts/images_fkcdn.zip'
+			link_tunics='https://transfer.sh/8gQj4/tunics.csv'
+			if os.path.exists('./dataset/images_fkcdn.zip'):
+				os.system('unzip ./dataset/images_fkcdn.zip')
+			else:
+				urlretrieve(link_zip,'./dataset/images_fkcdn.zip')
+				urlretrieve(link_tunics,'./dataset/tunics.csv')
+		except Exception as e:
+			print(e)
+
+	RecModel=init_model()
+	database=create_encodings()
+	df = pd.read_csv('tunics.csv')
+	productFamily=json.load(open('pf.json'))
+	imgs=sorted(os.listdir('./images'))
+	#Compare each image with all other image in array
+	checks={}
+	ptr=0
+
+	# This loops compares the image with the preceeding images and 
+	# not with the already visited images to overcome redundancy
+	for i in range(2):
+	    anchor_image=imgs[i]
+	    anchor_name=imgs[i].split('.')[0]
+	    checks[anchor_name]=[]
+	    for j in range(i+1,len(imgs)):
+	        target_image=imgs[j]
+	        target_name=imgs[j].split('.')[0]
+	        if target_name not in productFamily[anchor_name] and verify(anchor_image, target_name, database, RecModel)[1] :
+	            checks[anchor_name].append(target_name)
+	            print(ptr)
+	            ptr+=1
+
+    # Dump the json to the temp server
+    import json
+    with open('output.json','w') as f:
+        json.dump(checks,f)