facebook_posts_analysis.py

facebook_posts_analysis.py

import sys
import os
import numpy as np
import tensorflow as tf

sys.path.append("..")

from object_detection.utils import label_map_util


os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'

MODEL_NAME = 'faster_rcnn_inception_resnet_v2_atrous_coco_11_06_2017'

MODEL_FILE = MODEL_NAME + '.tar.gz'
DOWNLOAD_BASE = 'http://download.tensorflow.org/models/object_detection/'
PATH_TO_CKPT = MODEL_NAME + '/frozen_inference_graph.pb'
PATH_TO_LABELS = os.path.join('data', 'mscoco_label_map.pbtxt')
NUM_CLASSES = 90

label_map = label_map_util.load_labelmap(PATH_TO_LABELS)
categories = label_map_util.convert_label_map_to_categories(label_map, max_num_classes=NUM_CLASSES,
                                                            use_display_name=True)
category_index = label_map_util.create_category_index(categories)


def detect_alert(boxes, classes, scores, category_index, max_boxes_to_draw=20,
                 min_score_thresh=.5,
                 ):
    r = []
    for i in range(min(max_boxes_to_draw, boxes.shape[0])):
        if scores is None or scores[i] > min_score_thresh:
            test1 = None
            test2 = None

            if category_index[classes[i]]['name']:
                test1 = category_index[classes[i]]['name']
                test2 = int(100 * scores[i])

            line = {}
            line[test1] = test2
            r.append(line)

    return r


def detect_objects(image_np, sess, detection_graph):
    image_np_expanded = np.expand_dims(image_np, axis=0)
    image_tensor = detection_graph.get_tensor_by_name('image_tensor:0')

    boxes = detection_graph.get_tensor_by_name('detection_boxes:0')

    scores = detection_graph.get_tensor_by_name('detection_scores:0')
    classes = detection_graph.get_tensor_by_name('detection_classes:0')
    num_detections = detection_graph.get_tensor_by_name('num_detections:0')

    # Actual detection.
    (boxes, scores, classes, num_detections) = sess.run(
        [boxes, scores, classes, num_detections],
        feed_dict={image_tensor: image_np_expanded})

    alert_array = detect_alert(np.squeeze(boxes), np.squeeze(classes).astype(np.int32), np.squeeze(scores),
                               category_index)

    return alert_array

def load_image_into_numpy_array(image):
    (im_width, im_height) = image.size
    return np.array(image.getdata()).reshape(
        (im_height, im_width, 3)).astype(np.uint8)

detection_graph = tf.Graph()
with detection_graph.as_default():
    od_graph_def = tf.GraphDef()
    with tf.gfile.GFile(PATH_TO_CKPT, 'rb') as fid:
        serialized_graph = fid.read()
        od_graph_def.ParseFromString(serialized_graph)
        tf.import_graph_def(od_graph_def, name='')


def process_image(image):
    with detection_graph.as_default():
        with tf.Session(graph=detection_graph) as sess:
            alert_array = detect_objects(image, sess, detection_graph)
            return alert_array

import facebook
import urllib, cStringIO
from PIL import Image

def facebook_posts_analysis(username, num_post, token):

    access_token = token
    user = username

    graph = facebook.GraphAPI(access_token)
    profile = graph.get_object(user)

    posts = graph.get_connections(profile['id'], 'posts', limit=num_post)

    for p in posts['data']:
        postid = p['id']
        print(p['created_time'].encode('utf-8') + ";" + p['id'].encode('utf-8') + ";" + p['message'].encode('utf-8'))

        try:
            a = graph.get_connections(postid, 'attachments', limit=1)

            for att in a['data']:
                file = cStringIO.StringIO(urllib.urlopen(att['media']['image']['src']).read())
                img = Image.open(file)

                if img:
                    list_elements = process_image(img)
                    print(list_elements)

        except Exception:
            pass


facebook_user = '' # change
num_post = 10 #change
token = '' #change


facebook_posts_analysis(facebook_user, num_post, token)