embed_image.py

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
from PIL import Image #NEW 

import tensorflow as tf
import numpy as np
import argparse
import facenet
import os
import sys
import math
import sklearn
from PIL import Image

class ImageEmbedding():
    def __init__(self, args):
        tf.Graph().as_default()
        self.sess = tf.Session()

        # Load the model
        print('Loading feature extraction model')
        facenet.load_model(args["model"], self.sess)
        
        # Get input and output tensors
        self.images_placeholder = tf.get_default_graph().get_tensor_by_name("input:0")
        self.embeddings = tf.get_default_graph().get_tensor_by_name("embeddings:0")
        self.phase_train_placeholder = tf.get_default_graph().get_tensor_by_name("phase_train:0")
        self.embedding_size = self.embeddings.get_shape()[1]

    def converter(self, imag):
        #img = Image.open("shia.jpg")
        img = imag
        img = img.resize((160, 160))
        img = np.asarray(img)
        img = np.reshape(img, [1, 160, 160, 3])
        print(img.shape)
        return img
                    
    def embed(self, imag):
        # Run forward pass to calculate embeddings

        img = self.converter(imag)
        #NEW
        #https://www.npmjs.com/package/facenet Check it out for image saving
        print (img)
        print(type(img))
        print(type(imag))
        image = Image.fromarray(np.squeeze(img), 'RGB')
        image.save('my.png')
        image.show()
        #NEW
        print('Calculating features for images')
        feed_dict = { self.images_placeholder:img, self.phase_train_placeholder:False }
        emb = self.sess.run(self.embeddings, feed_dict=feed_dict)
        return emb        

imgemb = ImageEmbedding({
    "model": "20170512-110547/"
})

#print(imgemb.embed("shia.jpg"))