Image_Generation_From_Captions.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\sbpat\\AppData\\Local\\Continuum\\anaconda3\\envs\\DLENV\\lib\\site-packages\\h5py\\__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n"
     ]
    }
   ],
   "source": [
    "import json\n",
    "import os\n",
    "from os.path import join, isfile\n",
    "import re\n",
    "import numpy as np\n",
    "import pickle\n",
    "import argparse\n",
    "import skipthoughts\n",
    "import h5py\n",
    "import time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def save_caption_vectors(data_dir):\n",
    "    img_dir = join(data_dir, \"flowers/jpg\")\n",
    "    image_files = [f for f in os.listdir(img_dir) if \"jpg\" in f]\n",
    "    print(image_files[300:400])\n",
    "    print(len(image_files))\n",
    "    image_captions = {img_file: [] for img_file in image_files}\n",
    "    \n",
    "    caption_dir = join(data_dir, \"flowers/text_c10\")\n",
    "    class_dirs = []\n",
    "    for i in range(1,103):\n",
    "        class_dir_name = \"class_%.5d\"%(i)\n",
    "        class_dirs.append(join(caption_dir,class_dir_name))\n",
    "    \n",
    "    for class_dir in class_dirs:\n",
    "        caption_files = [f for f in os.listdir(class_dir) if \"txt\" in f]\n",
    "        for cap_file in caption_files:\n",
    "            with open(join(class_dir,cap_file)) as f:\n",
    "                captions = f.read().split(\"\\n\")\n",
    "            img_file = cap_file[0:11] + \".jpg\"\n",
    "            #5 captions per image\n",
    "            image_captions[img_file] += [cap for cap in captions if len(cap)>0][0:5]\n",
    "    \n",
    "    print(len(captions))\n",
    "    \n",
    "    model = skipthoughts.load_model()\n",
    "    encoded_captions = {}\n",
    "    for i,img in enumerate(image_captions):\n",
    "        st = time.time()\n",
    "        encoded_captions[img] = skipthoughts.encode(model,image_captions[img])\n",
    "        print(i,len(image_captions),img)\n",
    "        print(\"Seconds\", time.time() - st)\n",
    "        \n",
    "    h = h5py.File(join(data_dir,\"flower_tv.hdf5\"))\n",
    "    for key in encoded_captions:\n",
    "        h.create_dataset(key, data = encoded_captions[key])\n",
    "    h.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "save_caption_vectors('Data')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import ops\n",
    "\n",
    "class GAN:\n",
    "    def __init__(self,options):\n",
    "        self.options = options\n",
    "        self.g_bn0 = ops.batch_norm(name = \"g_bn0\")\n",
    "        self.g_bn1 = ops.batch_norm(name = \"g_bn1\")\n",
    "        self.g_bn2 = ops.batch_norm(name = \"g_bn2\")\n",
    "        self.g_bn3 = ops.batch_norm(name = \"g_bn3\")\n",
    "        \n",
    "        self.d_bn1 = ops.batch_norm(name = \"d_bn1\")\n",
    "        self.d_bn2 = ops.batch_norm(name = \"d_bn2\")\n",
    "        self.d_bn3 = ops.batch_norm(name = \"d_bn3\")\n",
    "        self.d_bn4 = ops.batch_norm(name = \"d_bn4\")\n",
    "    #DC-GAN generator    \n",
    "    def generator(self,t_z,t_text_embedding):\n",
    "        with tf.variable_scope(tf.get_variable_scope(),reuse = tf.AUTO_REUSE):\n",
    "            s = self.options[\"image_size\"]\n",
    "\n",
    "            s2, s4, s8, s16 = int(s/2), int(s/4), int(s/8), int(s/16)\n",
    "\n",
    "            reduced_text_embedding = ops.lrelu(ops.linear(t_text_embedding,self.options[\"t_dim\"],\"g_embedding\"))\n",
    "            z_concat = tf.concat( [t_z, reduced_text_embedding],1)\n",
    "            z_ = ops.linear(z_concat, self.options[\"gf_dim\"]*8*s16*s16, \"g_h0_lin\")\n",
    "            h0 = tf.reshape(z_, [-1,s16,s16, self.options[\"gf_dim\"]*8])\n",
    "            h0 = tf.nn.relu(self.g_bn0(h0))\n",
    "\n",
    "            h1 = ops.deconv2d(h0, [self.options[\"batch_size\"], s8, s8, self.options[\"gf_dim\"]*4],name = \"g_h1\")\n",
    "            h1 = tf.nn.relu(self.g_bn1(h1))\n",
    "\n",
    "            h2 = ops.deconv2d(h1, [self.options[\"batch_size\"], s4, s4, self.options[\"gf_dim\"]*2],name = \"g_h2\")\n",
    "            h2 = tf.nn.relu(self.g_bn2(h2))\n",
    "\n",
    "            h3 = ops.deconv2d(h2, [self.options[\"batch_size\"], s2, s2, self.options[\"gf_dim\"]*1],name = \"g_h3\")\n",
    "            h3 = tf.nn.relu(self.g_bn3(h3))\n",
    "\n",
    "            h4 = ops.deconv2d(h3, [self.options[\"batch_size\"], s, s, 3], name = \"g_h4\")\n",
    "\n",
    "        return (tf.tanh(h4)/2. + 0.5)\n",
    "    \n",
    "    #DC-GAN discriminator\n",
    "    def discriminator(self,image,t_text_embedding,reuse = True):\n",
    "        #64 -> 32\n",
    "        with tf.variable_scope(tf.get_variable_scope(), reuse = tf.AUTO_REUSE):\n",
    "            h0 = ops.lrelu(ops.conv2d(image,self.options[\"df_dim\"], name = \"d_h0_conv\"))\n",
    "            #32 -> 16\n",
    "            h1 = ops.lrelu(self.d_bn1(ops.conv2d(h0,self.options[\"df_dim\"]*2, name = \"d_h1_conv\")))\n",
    "            #16 -> 8\n",
    "            h2 = ops.lrelu(self.d_bn2(ops.conv2d(h1,self.options[\"df_dim\"]*4, name = \"d_h2_conv\")))\n",
    "            #8 -> 4\n",
    "            h3 = ops.lrelu(self.d_bn3(ops.conv2d(h2,self.options[\"df_dim\"]*8, name = \"d_h3_conv\")))\n",
    "\n",
    "            reduced_text_embeddings = ops.lrelu(ops.linear(t_text_embedding,self.options[\"t_dim\"],\"d_embedding\"))\n",
    "            reduced_text_embeddings = tf.expand_dims(reduced_text_embeddings, 1)\n",
    "            reduced_text_embeddings = tf.expand_dims(reduced_text_embeddings, 2)\n",
    "            tiled_embeddings = tf.tile(reduced_text_embeddings,[1,4,4,1], name = \"tiled_embeddings\")\n",
    "            h3_concat = tf.concat( [h3,tiled_embeddings],3, name = \"h3_concat\")\n",
    "            h3_new = ops.lrelu(self.d_bn4(ops.conv2d(h3_concat,self.options[\"df_dim\"]*8,1,1,1,1, name = \"d_h3_conv_new\")))\n",
    "\n",
    "            h4 = ops.linear(tf.reshape(h3_new, [ self.options[\"batch_size\"],-1]), 1, \"d_h3_lin\")\n",
    "        \n",
    "        return tf.nn.sigmoid(h4), h4\n",
    "        \n",
    "    def build_model(self):\n",
    "        img_size = self.options[\"image_size\"]\n",
    "        t_real_image = tf.placeholder(\"float32\",[self.options[\"batch_size\"],img_size,img_size,3],name=\"real_image\")\n",
    "        t_wrong_image = tf.placeholder(\"float32\",[self.options[\"batch_size\"],img_size,img_size,3],name=\"wrong_image\")\n",
    "        t_real_caption = tf.placeholder(\"float32\",[self.options[\"batch_size\"],self.options[\"caption_vector_length\"]],name = \"real_caption_input\")\n",
    "        t_z = tf.placeholder(\"float32\",[self.options[\"batch_size\"],self.options[\"z_dim\"]])\n",
    "        \n",
    "        fake_image = self.generator(t_z,t_real_caption)\n",
    "        \n",
    "        disc_real_image, disc_real_image_logits = self.discriminator(t_real_image,t_real_caption)\n",
    "        disc_wrong_image, disc_wrong_image_logits = self.discriminator(t_wrong_image,t_real_caption, reuse = True)\n",
    "        disc_fake_image, disc_fake_image_logits = self.discriminator(fake_image,t_real_caption,reuse = True)\n",
    "        \n",
    "        g_loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(logits=disc_fake_image_logits,labels= tf.ones_like(disc_fake_image)))\n",
    "        \n",
    "        d_loss1 = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(logits=disc_real_image_logits,labels=tf.ones_like(disc_real_image)))\n",
    "        d_loss2 = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(logits=disc_wrong_image_logits,labels=tf.zeros_like(disc_wrong_image)))\n",
    "        d_loss3 = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(logits=disc_fake_image_logits,labels=tf.zeros_like(disc_fake_image)))\n",
    "        \n",
    "        d_loss = d_loss1 + d_loss2 + d_loss3\n",
    "        t_vars = tf.trainable_variables()\n",
    "        d_vars = [var for var in t_vars if \"d_\" in var.name]\n",
    "        g_vars = [var for var in t_vars if \"g_\" in var.name]\n",
    "        \n",
    "        input_tensors = {\n",
    "            \"t_real_image\": t_real_image,\n",
    "            \"t_wrong_image\": t_wrong_image,\n",
    "            \"t_real_caption\": t_real_caption,\n",
    "            \"t_z\": t_z\n",
    "        }\n",
    "        variables = {\n",
    "            \"d_vars\": d_vars,\n",
    "            \"g_vars\": g_vars\n",
    "        }\n",
    "        loss = {\n",
    "            \"g_loss\":g_loss,\n",
    "            \"d_loss\": d_loss\n",
    "        }\n",
    "        outputs = {\"generator\": fake_image}\n",
    "        checks = {\n",
    "            \"d_loss1\": d_loss1,\n",
    "            \"d_loss2\": d_loss2,\n",
    "            \"d_loss3\": d_loss3,\n",
    "            \"disc_real_image_logits\": disc_real_image_logits,\n",
    "            \"disc_wrong_image_logits\": disc_wrong_image,\n",
    "            \"disc_fake_image_logits\": disc_fake_image_logits\n",
    "        }\n",
    "        return input_tensors, variables,loss,outputs,checks\n",
    "    \n",
    "    def build_generator(self):\n",
    "        img_size = self.options[\"image_size\"]\n",
    "        t_real_caption = tf.placeholder(\"float32\",[self.options[\"batch_size\"],self.options[\"caption_vector_length\"]],name = \"real_caption_input\")\n",
    "        t_z = tf.placeholder(\"float32\",[self.options[\"batch_size\"],self.options[\"z_dim\"]])\n",
    "        fake_image = self.sampler(t_z,t_real_caption)\n",
    "        \n",
    "        input_tensors = {\"t_real_caption\":t_real_caption,\"t_z\": t_z}\n",
    "        outputs = {\"generator\":fake_image}\n",
    "        return input_tensors,outputs\n",
    "    \n",
    "    def sampler(self,t_z, t_text_embedding):\n",
    "        with tf.variable_scope(tf.get_variable_scope(),reuse = tf.AUTO_REUSE):\n",
    "            s = self.options[\"image_size\"]\n",
    "            s2, s4, s8, s16 = int(s/2), int(s/4), int(s/8), int(s/16)\n",
    "\n",
    "            reduced_text_embedding = ops.lrelu(ops.linear(t_text_embedding,self.options[\"t_dim\"],\"g_embedding\"))\n",
    "            z_concat = tf.concat([t_z, reduced_text_embedding],1)\n",
    "            z_ = ops.linear(z_concat,self.options[\"gf_dim\"]*8*s16*s16,\"g_h0_lin\")\n",
    "\n",
    "            h0 = tf.reshape(z_,[-1,s16,s16,self.options[\"gf_dim\"]*8])\n",
    "            h0 = tf.nn.relu(self.g_bn0(h0,train = False))\n",
    "\n",
    "            h1 = ops.deconv2d(h0, [self.options[\"batch_size\"],s8,s8,self.options[\"gf_dim\"]*4],name = \"g_h1\")\n",
    "            h1 = tf.nn.relu(self.g_bn1(h1,train = False))\n",
    "\n",
    "            h2 = ops.deconv2d(h1, [self.options[\"batch_size\"],s4,s4,self.options[\"gf_dim\"]*2],name = \"g_h2\")\n",
    "            h2 = tf.nn.relu(self.g_bn2(h2,train = False))\n",
    "\n",
    "            h3 = ops.deconv2d(h2, [self.options[\"batch_size\"],s2,s2,self.options[\"gf_dim\"]*1],name = \"g_h3\")\n",
    "            h3 = tf.nn.relu(self.g_bn3(h3,train = False))\n",
    "\n",
    "            h4 = ops.deconv2d(h3, [self.options[\"batch_size\"],s,s,3], name = \"g_h4\")\n",
    "        \n",
    "        return (tf.tanh(h4)/2. + 0.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import scipy.misc\n",
    "import image_processing\n",
    "import random\n",
    "import shutil\n",
    "import imageio\n",
    "from skimage import img_as_ubyte"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def main():\n",
    "    z_dim = 100\n",
    "    t_dim = 256\n",
    "    batch_size = 256\n",
    "    image_size = 64\n",
    "    gf_dim = 64\n",
    "    df_dim = 64\n",
    "    gfc_dim = 1024\n",
    "    caption_vector_length = 2400\n",
    "    data_dir = \"Data\"\n",
    "    learning_rate = 0.0003\n",
    "    beta1 = 0.5\n",
    "    epochs = 200\n",
    "    save_every = 20\n",
    "    resume_model = True\n",
    "    data_set = \"flowers\"\n",
    "    model_options = {\n",
    "        \"z_dim\" : z_dim,\n",
    "        \"t_dim\" : t_dim,\n",
    "        \"batch_size\" : batch_size,\n",
    "        \"image_size\" : image_size,\n",
    "        \"gf_dim\" : gf_dim,\n",
    "        \"df_dim\" : df_dim,\n",
    "        \"gfc_dim\" : gfc_dim,\n",
    "        \"caption_vector_length\" : caption_vector_length\n",
    "    }\n",
    "    gan = GAN(model_options)\n",
    "    input_tensors,variables,loss,outputs,checks = gan.build_model()\n",
    "    d_optim = tf.train.AdamOptimizer(learning_rate,beta1 = beta1).minimize(loss[\"d_loss\"],var_list = variables[\"d_vars\"])\n",
    "    g_optim = tf.train.AdamOptimizer(learning_rate,beta1 = beta1).minimize(loss[\"g_loss\"],var_list = variables[\"g_vars\"])\n",
    "    \n",
    "    sess = tf.InteractiveSession()\n",
    "    tf.initialize_all_variables().run()\n",
    "    \n",
    "    saver = tf.train.Saver()\n",
    "    if resume_model:\n",
    "        saver = tf.train.import_meta_graph('Data/Models/latest_mdoel_flowers_temp.ckpt.index')\n",
    "        saver.restore(sess, \"Data/Models/latest_mdoel_flowers_temp.ckpt\")\n",
    "        \n",
    "    loaded_data = load_training_data(data_dir, data_set)\n",
    "    \n",
    "    for i in range(epochs):\n",
    "        batch_no = 0\n",
    "        while batch_no*batch_size < loaded_data[\"data_length\"]:\n",
    "            real_images,wrong_images,caption_vectors,z_noise,image_files = get_training_batch(batch_no,batch_size,\n",
    "                image_size,z_dim,caption_vector_length,\"train\",data_dir,data_set, loaded_data)\n",
    "            \n",
    "            #Update Discriminator\n",
    "            check_ts = [checks[\"d_loss1\"], checks[\"d_loss2\"], checks[\"d_loss3\"]]\n",
    "            _, d_loss, gen,d1,d2,d3 = sess.run([d_optim, loss[\"d_loss\"], outputs[\"generator\"]]+check_ts,\n",
    "                                              feed_dict={\n",
    "                                                  input_tensors[\"t_real_image\"]: real_images,\n",
    "                                                  input_tensors[\"t_wrong_image\"]: wrong_images,\n",
    "                                                  input_tensors[\"t_real_caption\"]: caption_vectors,\n",
    "                                                  input_tensors[\"t_z\"]: z_noise,\n",
    "                                              })\n",
    "            \n",
    "            print(\"d1\",d1)\n",
    "            print(\"d2\",d2)\n",
    "            print(\"d3\",d3)\n",
    "            print(\"D\",d_loss)\n",
    "            #Update Generator\n",
    "            _, g_loss, gen = sess.run([g_optim, loss[\"g_loss\"], outputs[\"generator\"]],\n",
    "                feed_dict = {\n",
    "                    input_tensors[\"t_real_image\"] : real_images,\n",
    "                    input_tensors[\"t_wrong_image\"] : wrong_images,\n",
    "                    input_tensors[\"t_real_caption\"] : caption_vectors,\n",
    "                    input_tensors[\"t_z\"] : z_noise,\n",
    "                })\n",
    "            # GEN UPDATE TWICE, to make sure d_loss does not go to 0\n",
    "            _, g_loss, gen = sess.run([g_optim, loss[\"g_loss\"], outputs[\"generator\"]],\n",
    "                feed_dict = {\n",
    "                    input_tensors[\"t_real_image\"] : real_images,\n",
    "                    input_tensors[\"t_wrong_image\"] : wrong_images,\n",
    "                    input_tensors[\"t_real_caption\"] : caption_vectors,\n",
    "                    input_tensors[\"t_z\"] : z_noise,\n",
    "                })\n",
    "            print(\"LOSSES\",d_loss,g_loss,batch_no,i,len(loaded_data[\"image_list\"])/batch_size)\n",
    "            batch_no +=1\n",
    "            if (batch_no%save_every) == 0:\n",
    "                print(\"Saving Images, Model\")\n",
    "                save_for_vis(data_dir,real_images,gen,image_files)\n",
    "                save_path = saver.save(sess,\"Data/Models/latest_mdoel_{}_temp.ckpt\".format(data_set))\n",
    "        if i%5==0:\n",
    "            save_path = saver.save(sess,\"Data/Models/model_after_{}_epoch_{}.ckpt\".format(data_set,i))\n",
    "            \n",
    "def load_training_data(data_dir, data_set):\n",
    "    if data_set == \"flowers\":\n",
    "        h = h5py.File(join(data_dir, \"flower_tv.hdf5\"))\n",
    "        flower_captions = {}\n",
    "        for ds in h.items():\n",
    "            flower_captions[ds[0]] = np.array(ds[1])\n",
    "        image_list = [key for key in flower_captions]\n",
    "        image_list.sort()\n",
    "\n",
    "        img_75 = int(len(image_list)*0.75)\n",
    "        training_image_list = image_list[0:img_75]\n",
    "        random.shuffle(training_image_list)\n",
    "        \n",
    "        return {\n",
    "            \"image_list\" : training_image_list,\n",
    "            \"captions\" : flower_captions,\n",
    "            \"data_length\" : len(training_image_list)\n",
    "        }\n",
    "    \n",
    "    else:\n",
    "        with open(join(data_dir, \"meta_train.pkl\")) as f:\n",
    "            meta_data = pickle.load(f)\n",
    "        # No preloading for MS-COCO\n",
    "        return meta_data\n",
    "            \n",
    "def save_for_vis(data_dir, real_images, generated_images, image_files):\n",
    "    #create a set of 10 random numbers to save 10 random images in samples directory\n",
    "    a= set()\n",
    "    while len(a)<=10:\n",
    "        a.update(random.sample(range(*sorted([0,real_images.shape[0]])),10))\n",
    "    \n",
    "    for i in list(a)[:10]:\n",
    "        real_image_255 = np.zeros( (64,64,3), dtype=np.uint8)\n",
    "        real_images_255 = img_as_ubyte((real_images[i,:,:,:]))\n",
    "        imageio.imwrite( join(data_dir, \"samples/{}_{}.jpg\".format(i, image_files[i].split(\"/\")[-1] )) , real_images_255)\n",
    "\n",
    "        fake_image_255 = np.zeros( (64,64,3), dtype=np.uint8)\n",
    "        fake_images_255 = img_as_ubyte((generated_images[i,:,:,:]))\n",
    "        imageio.imwrite(join(data_dir, \"samples/fake_image_{}.jpg\".format(i)), fake_images_255)\n",
    "        \n",
    "def get_training_batch(batch_no, batch_size, image_size, z_dim, \n",
    "    caption_vector_length, split, data_dir, data_set, loaded_data = None):\n",
    "    if data_set == \"mscoco\":\n",
    "        with h5py.File( join(data_dir, \"tvs/\"+split + \"_tvs_\" + str(batch_no))) as hf:\n",
    "            caption_vectors = np.array(hf.get(\"tv\"))\n",
    "            caption_vectors = caption_vectors[:,0:caption_vector_length]\n",
    "        with h5py.File( join(data_dir, \"tvs/\"+split + \"_tv_image_id_\" + str(batch_no))) as hf:\n",
    "            image_ids = np.array(hf.get(\"tv\"))\n",
    "\n",
    "        real_images = np.zeros((batch_size, 64, 64, 3))\n",
    "        wrong_images = np.zeros((batch_size, 64, 64, 3))\n",
    "        \n",
    "        image_files = []\n",
    "        for idx, image_id in enumerate(image_ids):\n",
    "            image_file = join(data_dir, \"%s2014/COCO_%s2014_%.12d.jpg\"%(split, split, image_id) )\n",
    "            image_array = image_processing.load_image_array(image_file, image_size)\n",
    "            real_images[idx,:,:,:] = image_array\n",
    "            image_files.append(image_file)\n",
    "        \n",
    "        # TODO>> As of Now, wrong images are just shuffled real images.\n",
    "        first_image = real_images[0,:,:,:]\n",
    "        for i in range(0, batch_size):\n",
    "            if i < batch_size - 1:\n",
    "                wrong_images[i,:,:,:] = real_images[i+1,:,:,:]\n",
    "            else:\n",
    "                wrong_images[i,:,:,:] = first_image\n",
    "        z_noise = np.random.uniform(-1, 1, [batch_size, z_dim])\n",
    "\n",
    "\n",
    "        return real_images, wrong_images, caption_vectors, z_noise, image_files\n",
    "\n",
    "    if data_set == \"flowers\":\n",
    "        real_images = np.zeros((batch_size, 64, 64, 3))\n",
    "        wrong_images = np.zeros((batch_size, 64, 64, 3))\n",
    "        captions = np.zeros((batch_size, caption_vector_length))\n",
    "\n",
    "        cnt = 0\n",
    "        image_files = []\n",
    "        for i in range(batch_no * batch_size, batch_no * batch_size + batch_size):\n",
    "            idx = i % len(loaded_data[\"image_list\"])\n",
    "            image_file =  join(data_dir, \"flowers/jpg/\"+loaded_data[\"image_list\"][idx])\n",
    "            image_array = image_processing.load_image_array(image_file, image_size)\n",
    "            real_images[cnt,:,:,:] = image_array\n",
    "            \n",
    "            # Improve this selection of wrong image\n",
    "            wrong_image_id = random.randint(0,len(loaded_data[\"image_list\"])-1)\n",
    "            wrong_image_file =  join(data_dir, \"flowers/jpg/\"+loaded_data[\"image_list\"][wrong_image_id])\n",
    "            wrong_image_array = image_processing.load_image_array(wrong_image_file, image_size)\n",
    "            wrong_images[cnt, :,:,:] = wrong_image_array\n",
    "\n",
    "            random_caption = random.randint(0,4)\n",
    "            captions[cnt,:] = loaded_data[\"captions\"][ loaded_data[\"image_list\"][idx] ][ random_caption ][0:caption_vector_length]\n",
    "            image_files.append( image_file )\n",
    "            cnt += 1\n",
    "\n",
    "        z_noise = np.random.uniform(-1, 1, [batch_size, z_dim])\n",
    "        return real_images, wrong_images, captions, z_noise, image_files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "main()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def generate_thought_vectors():\n",
    "    caption_file = \"Data/sample_captions.txt\"\n",
    "    data_dir = \"Data\"\n",
    "    with open(caption_file) as f:\n",
    "        captions = f.read().split(\"\\n\")\n",
    "    captions = [cap for cap in captions if len(cap)>0]\n",
    "    print(captions)\n",
    "    model = skipthoughts.load_model()\n",
    "    caption_vectors = skipthoughts.encode(model,captions)\n",
    "    if os.path.isfile(join(data_dir, \"sample_caption_vectors.hdf5\")):\n",
    "        os.remove(join(data_dir,\"sample_caption_vectors.hdf5\"))\n",
    "    h = h5py.File(join(data_dir,\"sample_caption_vectors.hdf5\"))\n",
    "    h.create_dataset(\"vectors\", data = caption_vectors)\n",
    "    h.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "generate_thought_vectors()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "import model\n",
    "import ops\n",
    "import cv2\n",
    "import PIL"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "def generate_images():\n",
    "    z_dim = 100\n",
    "    t_dim = 256\n",
    "    image_size = 64\n",
    "    gf_dim = 64\n",
    "    df_dim = 64\n",
    "    gfc_dim = 1024\n",
    "    caption_vector_length = 2400\n",
    "    data_dir = \"Data\"\n",
    "    model_path = \"Data/Models/latest_model_flowers_temp.ckpt\"\n",
    "    n_images = 400\n",
    "    caption_thought_vectors = \"Data/sample_caption_vectors.hdf5\"\n",
    "    model_options = {\n",
    "        \"z_dim\" : z_dim,\n",
    "        \"t_dim\" : t_dim,\n",
    "        \"batch_size\" : n_images,\n",
    "        \"image_size\" : image_size,\n",
    "        \"gf_dim\" : gf_dim,\n",
    "        \"df_dim\" : df_dim,\n",
    "        \"gfc_dim\" : gfc_dim,\n",
    "        \"caption_vector_length\" : caption_vector_length\n",
    "    }\n",
    "    gan = GAN(model_options)\n",
    "    _,_,_,_,_ = gan.build_model()\n",
    "    sess = tf.InteractiveSession()\n",
    "    \n",
    "#     saver = tf.train.Saver()\n",
    "#     saver.restore(sess,model_path)\n",
    "    saver = tf.train.import_meta_graph('Data/Models/latest_mdoel_flowers_temp.ckpt.index')\n",
    "    saver.restore(sess, \"Data/Models/latest_mdoel_flowers_temp.ckpt\")\n",
    "    \n",
    "    input_tensors, outputs = gan.build_generator()\n",
    "    \n",
    "    h = h5py.File(caption_thought_vectors)\n",
    "    caption_vectors = np.array(h[\"vectors\"])\n",
    "    caption_image_dic = {}\n",
    "    for cn, caption_vectors in enumerate(caption_vectors):\n",
    "        caption_images = []\n",
    "        z_noise = np.random.uniform(-1,1,[n_images, z_dim])\n",
    "        caption = [caption_vectors[0:caption_vector_length]]*n_images\n",
    "        [gen_image] = sess.run([outputs[\"generator\"]],\n",
    "            feed_dict = {\n",
    "                input_tensors[\"t_real_caption\"]: caption,\n",
    "                input_tensors[\"t_z\"]: z_noise,\n",
    "            })\n",
    "        caption_images = [gen_image[i,:,:,:] for i in range(0,n_images)]\n",
    "        caption_image_dic[cn] = caption_images\n",
    "        print(\"Generated\", cn)\n",
    "    for f in os.listdir(join(data_dir, \"val_samples\")):\n",
    "        if os.path.isfile(f):\n",
    "            os.unlink(join(data_dir,\"val_samples/\"+f))\n",
    "    \n",
    "    for cn in range(0, len(caption_vectors)-1):\n",
    "        caption_images = []\n",
    "        for i, im in enumerate( caption_image_dic[ cn ] ):\n",
    "            # im_name = \"caption_{}_{}.jpg\".format(cn, i)\n",
    "            # scipy.misc.imsave( join(args.data_dir, 'val_samples/{}'.format(im_name)) , im)\n",
    "            caption_images.append( im )\n",
    "            scipy.misc.imsave( join(data_dir, 'val_samples/combined_image_%d_%04d.jpg' %(cn,i)) , im)\n",
    "        caption_images = np.array(caption_images)\n",
    "        caption_images = (caption_images*255).astype(np.uint8)\n",
    "        make_video(caption_images)\n",
    "        \n",
    "            \n",
    "#     for cn in range(0, len(caption_vectors)):\n",
    "#         caption_images = []\n",
    "#         for i, im in enumerate( caption_image_dic[ cn ] ):\n",
    "#             caption_images.append( im )\n",
    "            \n",
    "#         make_video(caption_images)\n",
    "#         for  i in range(len(caption_images)):\n",
    "#             imageio.imwrite(join(data_dir, \"val_samples/{}.jpg\".format(i)), np.uint8(caption_images[i]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Restoring parameters from Data/Models/latest_mdoel_flowers_temp.ckpt\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\sbpat\\AppData\\Local\\Continuum\\anaconda3\\envs\\DLENV\\lib\\site-packages\\tensorflow\\python\\framework\\meta_graph.py:637: RuntimeWarning: Unexpected end-group tag: Not all data was converted\n",
      "  meta_graph_def.ParseFromString(file_content)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Generated 0\n",
      "Generated 1\n",
      "Generated 2\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\sbpat\\AppData\\Local\\Continuum\\anaconda3\\envs\\DLENV\\lib\\site-packages\\ipykernel_launcher.py:59: DeprecationWarning: `imsave` is deprecated!\n",
      "`imsave` is deprecated in SciPy 1.0.0, and will be removed in 1.2.0.\n",
      "Use ``imageio.imwrite`` instead.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[[[1.55701637e-01 1.95370972e-01 1.17713958e-01]\n",
      "   [9.70563293e-02 1.17971867e-01 8.81196260e-02]\n",
      "   [1.08201563e-01 8.74434114e-02 1.10710263e-01]\n",
      "   ...\n",
      "   [2.15970486e-01 2.30552971e-01 3.01989853e-01]\n",
      "   [2.18476713e-01 2.22276211e-01 3.13048124e-01]\n",
      "   [2.45469451e-01 2.82682091e-01 3.06617141e-01]]\n",
      "\n",
      "  [[8.30275118e-02 1.22683585e-01 6.46986663e-02]\n",
      "   [1.11763656e-01 9.35676694e-02 7.38447011e-02]\n",
      "   [1.56059355e-01 1.13070339e-01 1.18269622e-01]\n",
      "   ...\n",
      "   [2.25871563e-01 2.27404714e-01 2.71698892e-01]\n",
      "   [3.20263445e-01 3.30670834e-01 3.71549308e-01]\n",
      "   [3.08630288e-01 3.27499062e-01 3.74122441e-01]]\n",
      "\n",
      "  [[6.35487437e-02 1.06324434e-01 5.85293174e-02]\n",
      "   [1.60197467e-01 1.26789629e-01 1.35137618e-01]\n",
      "   [1.18559748e-01 1.12850845e-01 1.34008139e-01]\n",
      "   ...\n",
      "   [2.75586784e-01 2.50427544e-01 2.13617921e-01]\n",
      "   [2.53101289e-01 3.29911172e-01 2.76510537e-01]\n",
      "   [2.86723167e-01 3.18481147e-01 3.07479739e-01]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[8.49150419e-02 1.22905910e-01 2.03258514e-01]\n",
      "   [1.22577429e-01 1.46897435e-01 2.31966883e-01]\n",
      "   [1.17791921e-01 1.45887315e-01 2.35728949e-01]\n",
      "   ...\n",
      "   [5.74808061e-01 5.31157970e-01 5.53151131e-01]\n",
      "   [5.96272230e-01 5.68230569e-01 7.07284331e-01]\n",
      "   [5.53861201e-01 5.16576469e-01 6.38962269e-01]]\n",
      "\n",
      "  [[8.60985518e-02 1.25368595e-01 1.98188573e-01]\n",
      "   [1.40984148e-01 1.60835743e-01 2.71868199e-01]\n",
      "   [1.11299157e-01 1.19844288e-01 2.12864965e-01]\n",
      "   ...\n",
      "   [6.22916341e-01 5.42450786e-01 6.65476441e-01]\n",
      "   [4.88812238e-01 5.61571717e-01 6.71892166e-01]\n",
      "   [4.33539867e-01 4.43922400e-01 5.56560755e-01]]\n",
      "\n",
      "  [[1.19153738e-01 1.28693044e-01 1.85180187e-01]\n",
      "   [9.72129703e-02 1.03928179e-01 1.84994459e-01]\n",
      "   [1.14996225e-01 1.35666788e-01 1.71418607e-01]\n",
      "   ...\n",
      "   [5.05080819e-01 4.37645942e-01 5.13504982e-01]\n",
      "   [2.77831256e-01 3.40104222e-01 4.03031915e-01]\n",
      "   [2.81316042e-01 2.99726665e-01 3.46014053e-01]]]\n",
      "\n",
      "\n",
      " [[[1.99119985e-01 2.51106828e-01 1.28605634e-01]\n",
      "   [1.40693307e-01 1.65943235e-01 1.01569057e-01]\n",
      "   [1.31344020e-01 1.44667417e-01 1.27771676e-01]\n",
      "   ...\n",
      "   [1.63345277e-01 1.59757882e-01 1.67927355e-01]\n",
      "   [1.80146933e-01 1.67202353e-01 1.55111253e-01]\n",
      "   [1.75837278e-01 2.05277175e-01 1.43200785e-01]]\n",
      "\n",
      "  [[1.29768729e-01 2.12464213e-01 8.85333419e-02]\n",
      "   [1.50871277e-01 1.52135849e-01 9.17928517e-02]\n",
      "   [1.07887715e-01 1.20038778e-01 8.88661742e-02]\n",
      "   ...\n",
      "   [2.05386758e-01 1.71179444e-01 1.37965083e-01]\n",
      "   [3.16495210e-01 2.65392810e-01 2.18019307e-01]\n",
      "   [2.64701903e-01 2.48189986e-01 2.02043235e-01]]\n",
      "\n",
      "  [[5.68297803e-02 1.26561493e-01 4.67588603e-02]\n",
      "   [1.62823170e-01 1.54038846e-01 1.10984117e-01]\n",
      "   [1.07609689e-01 1.34451568e-01 1.15046114e-01]\n",
      "   ...\n",
      "   [2.45032966e-01 1.77965909e-01 1.44546896e-01]\n",
      "   [2.13333845e-01 2.31895775e-01 1.71662569e-01]\n",
      "   [2.75571346e-01 2.48345792e-01 1.98881269e-01]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[1.38514102e-01 1.71210408e-01 1.53179526e-01]\n",
      "   [1.80799425e-01 1.94196403e-01 1.53604656e-01]\n",
      "   [2.15965867e-01 2.12354004e-01 1.92068219e-01]\n",
      "   ...\n",
      "   [6.23495579e-01 6.11912727e-01 5.48857570e-01]\n",
      "   [5.96252382e-01 6.29630089e-01 6.18323505e-01]\n",
      "   [5.63497484e-01 5.18083632e-01 5.51469922e-01]]\n",
      "\n",
      "  [[1.62072390e-01 2.04747021e-01 1.91736162e-01]\n",
      "   [2.85297334e-01 2.48528838e-01 2.27693796e-01]\n",
      "   [2.09087431e-01 2.09364831e-01 1.93741858e-01]\n",
      "   ...\n",
      "   [5.86666346e-01 5.64481974e-01 5.14189065e-01]\n",
      "   [5.81316352e-01 6.19553566e-01 5.94939590e-01]\n",
      "   [5.13892353e-01 4.69993681e-01 4.66329783e-01]]\n",
      "\n",
      "  [[1.59886837e-01 1.95630848e-01 1.71250582e-01]\n",
      "   [1.77359730e-01 1.74841642e-01 1.77947164e-01]\n",
      "   [2.19224334e-01 2.33457208e-01 1.99114740e-01]\n",
      "   ...\n",
      "   [4.70553607e-01 4.21839714e-01 4.31422085e-01]\n",
      "   [4.10401493e-01 4.50198442e-01 3.95399272e-01]\n",
      "   [4.94047582e-01 4.70295578e-01 4.45022970e-01]]]\n",
      "\n",
      "\n",
      " [[[8.24723542e-02 1.28528088e-01 7.23557174e-02]\n",
      "   [4.24109399e-02 7.53407478e-02 4.47187424e-02]\n",
      "   [5.87216914e-02 5.53176403e-02 7.86389112e-02]\n",
      "   ...\n",
      "   [8.33954513e-02 1.14259332e-01 1.92873091e-01]\n",
      "   [7.20877647e-02 9.62752402e-02 1.66089773e-01]\n",
      "   [1.22218281e-01 1.77522928e-01 2.09761620e-01]]\n",
      "\n",
      "  [[3.48881483e-02 7.52619505e-02 3.98236513e-02]\n",
      "   [4.37318087e-02 4.87325788e-02 3.49275768e-02]\n",
      "   [6.66023195e-02 6.39663637e-02 6.83869123e-02]\n",
      "   ...\n",
      "   [1.35074496e-01 1.52760267e-01 2.50811517e-01]\n",
      "   [1.94241822e-01 2.50068545e-01 3.19173664e-01]\n",
      "   [1.54952288e-01 2.04147041e-01 2.55905956e-01]]\n",
      "\n",
      "  [[1.93884373e-02 4.42405939e-02 2.69584954e-02]\n",
      "   [5.78972101e-02 6.41696751e-02 7.41589665e-02]\n",
      "   [4.87863123e-02 5.99872470e-02 7.25284219e-02]\n",
      "   ...\n",
      "   [1.44518197e-01 1.82057381e-01 2.29434967e-01]\n",
      "   [1.13014728e-01 1.97445810e-01 2.26131171e-01]\n",
      "   [1.86836690e-01 2.48077035e-01 2.73101270e-01]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[5.81172705e-02 9.75008011e-02 2.11301982e-01]\n",
      "   [7.01377094e-02 1.08311474e-01 1.55686289e-01]\n",
      "   [7.31840134e-02 1.39749587e-01 1.28968060e-01]\n",
      "   ...\n",
      "   [2.46386349e-01 3.94778669e-01 4.46913093e-01]\n",
      "   [2.14923143e-01 4.75453317e-01 6.07636034e-01]\n",
      "   [2.61516333e-01 4.28966641e-01 5.27001083e-01]]\n",
      "\n",
      "  [[3.20551693e-02 7.21578002e-02 1.18768275e-01]\n",
      "   [6.26507401e-02 1.14242405e-01 2.02129722e-01]\n",
      "   [5.93579113e-02 1.27166271e-01 1.69308215e-01]\n",
      "   ...\n",
      "   [2.23719358e-01 4.29257154e-01 5.30058324e-01]\n",
      "   [9.52377319e-02 3.06770861e-01 3.96046370e-01]\n",
      "   [1.36700422e-01 3.57842684e-01 3.40114683e-01]]\n",
      "\n",
      "  [[5.42146862e-02 7.15469122e-02 1.26895428e-01]\n",
      "   [5.06263673e-02 8.19134712e-02 1.20667875e-01]\n",
      "   [9.26057994e-02 1.48065925e-01 1.90866053e-01]\n",
      "   ...\n",
      "   [1.39164746e-01 2.34767109e-01 3.38010192e-01]\n",
      "   [8.38589072e-02 2.31794506e-01 3.02712858e-01]\n",
      "   [6.38827682e-02 1.71080530e-01 2.28845179e-01]]]\n",
      "\n",
      "\n",
      " ...\n",
      "\n",
      "\n",
      " [[[1.36562645e-01 1.20886654e-01 4.36011851e-02]\n",
      "   [5.58260977e-02 6.47855699e-02 2.59117484e-02]\n",
      "   [6.16589487e-02 3.35305631e-02 2.36665606e-02]\n",
      "   ...\n",
      "   [5.00532389e-02 6.71867728e-02 5.72591424e-02]\n",
      "   [5.34038246e-02 5.55301905e-02 5.97300529e-02]\n",
      "   [6.98690116e-02 9.96459126e-02 6.98041320e-02]]\n",
      "\n",
      "  [[6.10385239e-02 8.78959000e-02 2.13801861e-02]\n",
      "   [2.71585882e-02 3.50760520e-02 1.05315745e-02]\n",
      "   [2.08318830e-02 1.60716176e-02 1.20763481e-02]\n",
      "   ...\n",
      "   [9.07491148e-02 9.39799845e-02 6.28094077e-02]\n",
      "   [1.18001670e-01 1.35237634e-01 7.75354505e-02]\n",
      "   [6.80819154e-02 8.35196376e-02 4.73689139e-02]]\n",
      "\n",
      "  [[1.43033266e-02 2.55721807e-02 8.02564621e-03]\n",
      "   [1.03639066e-02 2.17599273e-02 8.29526782e-03]\n",
      "   [3.25741172e-02 3.34681869e-02 1.64261460e-02]\n",
      "   ...\n",
      "   [1.19218647e-01 1.08494997e-01 7.84569979e-02]\n",
      "   [8.48292410e-02 1.06570214e-01 6.13054037e-02]\n",
      "   [7.46659636e-02 1.15575224e-01 6.34544194e-02]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[4.00743484e-02 6.03244305e-02 6.67798519e-02]\n",
      "   [1.90489888e-02 2.34079957e-02 2.14886963e-02]\n",
      "   [1.72103047e-02 2.50093341e-02 2.20194459e-02]\n",
      "   ...\n",
      "   [1.06522441e-03 1.96504593e-03 6.27517700e-04]\n",
      "   [1.26013160e-03 8.31577182e-03 1.90934539e-03]\n",
      "   [3.25793028e-03 5.27206063e-03 1.22305751e-03]]\n",
      "\n",
      "  [[2.93807387e-02 4.62712049e-02 5.32503426e-02]\n",
      "   [5.41880727e-02 7.31444955e-02 9.31284130e-02]\n",
      "   [2.85825729e-02 3.40654552e-02 3.94997001e-02]\n",
      "   ...\n",
      "   [2.70816684e-03 1.85219646e-02 6.36318326e-03]\n",
      "   [8.04662704e-05 4.85152006e-04 4.03732061e-04]\n",
      "   [1.19715929e-04 1.07359886e-03 1.54495239e-04]]\n",
      "\n",
      "  [[5.45755923e-02 7.10152090e-02 7.51663744e-02]\n",
      "   [4.50421274e-02 4.83374596e-02 6.67672753e-02]\n",
      "   [2.53996849e-02 3.51195931e-02 2.30787396e-02]\n",
      "   ...\n",
      "   [1.98298693e-03 1.36893690e-02 1.10756457e-02]\n",
      "   [6.92516565e-04 3.16086411e-03 1.02427602e-03]\n",
      "   [3.78251076e-04 2.17095017e-03 9.30607319e-04]]]\n",
      "\n",
      "\n",
      " [[[3.72021317e-01 3.57344687e-01 1.68460071e-01]\n",
      "   [3.08101922e-01 2.93899059e-01 1.42250508e-01]\n",
      "   [2.96946287e-01 2.72516817e-01 1.67094827e-01]\n",
      "   ...\n",
      "   [4.34289277e-02 8.14127922e-02 3.36772203e-02]\n",
      "   [6.54034019e-02 9.11215544e-02 5.69875836e-02]\n",
      "   [7.15931654e-02 1.18584991e-01 5.76456189e-02]]\n",
      "\n",
      "  [[2.80512631e-01 3.05253416e-01 9.59413052e-02]\n",
      "   [3.64728451e-01 3.05326730e-01 1.36425495e-01]\n",
      "   [3.95462841e-01 3.67052495e-01 2.28147447e-01]\n",
      "   ...\n",
      "   [7.11322427e-02 1.01422757e-01 3.24743688e-02]\n",
      "   [1.30596608e-01 1.70416921e-01 6.52492046e-02]\n",
      "   [7.75744915e-02 1.09284669e-01 4.20644879e-02]]\n",
      "\n",
      "  [[2.33908713e-01 2.63179660e-01 7.26763606e-02]\n",
      "   [4.63757157e-01 3.70169401e-01 1.78368270e-01]\n",
      "   [3.64543200e-01 3.90712857e-01 2.31200039e-01]\n",
      "   ...\n",
      "   [9.23579931e-02 1.18708342e-01 4.99382615e-02]\n",
      "   [7.19077885e-02 1.16279542e-01 3.84019911e-02]\n",
      "   [1.14693046e-01 1.53846383e-01 5.62154353e-02]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[2.46661007e-01 2.85734594e-01 1.99407637e-01]\n",
      "   [2.80266464e-01 2.91779190e-01 2.23553300e-01]\n",
      "   [2.86741316e-01 3.06964815e-01 2.62534618e-01]\n",
      "   ...\n",
      "   [4.34825271e-01 3.32849741e-01 2.88083643e-01]\n",
      "   [4.80152488e-01 4.25966203e-01 4.27698284e-01]\n",
      "   [4.70372766e-01 3.77620429e-01 3.42406631e-01]]\n",
      "\n",
      "  [[2.57638156e-01 3.02950740e-01 2.08763421e-01]\n",
      "   [3.32395792e-01 3.21915627e-01 2.32460797e-01]\n",
      "   [3.47733170e-01 3.63708854e-01 2.86176622e-01]\n",
      "   ...\n",
      "   [4.77276266e-01 3.36892307e-01 3.79770428e-01]\n",
      "   [2.48646051e-01 2.98871934e-01 2.68841892e-01]\n",
      "   [2.69620657e-01 2.95011699e-01 1.76353514e-01]]\n",
      "\n",
      "  [[2.40596265e-01 2.64187694e-01 1.96238935e-01]\n",
      "   [2.66552269e-01 2.57729828e-01 2.21527398e-01]\n",
      "   [3.40153456e-01 3.66232276e-01 2.91853130e-01]\n",
      "   ...\n",
      "   [1.53275788e-01 1.08545363e-01 7.87936747e-02]\n",
      "   [9.53431726e-02 1.10157132e-01 5.02296686e-02]\n",
      "   [9.42125320e-02 1.20519876e-01 5.92080355e-02]]]\n",
      "\n",
      "\n",
      " [[[2.01045573e-01 2.83579350e-01 1.63853228e-01]\n",
      "   [1.03505403e-01 1.18641466e-01 4.26592827e-02]\n",
      "   [2.60781467e-01 2.14209497e-01 1.88030899e-01]\n",
      "   ...\n",
      "   [7.37550259e-02 1.64414346e-01 2.11965799e-01]\n",
      "   [9.87893641e-02 1.45413995e-01 2.42541939e-01]\n",
      "   [8.67242217e-02 2.40800679e-01 2.74287194e-01]]\n",
      "\n",
      "  [[3.48085642e-01 5.12292922e-01 2.26076961e-01]\n",
      "   [8.14960301e-02 7.23983049e-02 3.81434560e-02]\n",
      "   [4.94505852e-01 4.66338754e-01 2.06131250e-01]\n",
      "   ...\n",
      "   [2.62402296e-02 3.09614837e-02 1.64665282e-02]\n",
      "   [2.66718984e-01 2.61929274e-01 2.43577033e-01]\n",
      "   [5.24494350e-02 1.04026020e-01 7.33925998e-02]]\n",
      "\n",
      "  [[1.93825096e-01 3.22359383e-01 1.18633717e-01]\n",
      "   [4.04449701e-02 6.15303218e-02 2.63518095e-02]\n",
      "   [2.60786712e-01 4.03743744e-01 2.91558653e-01]\n",
      "   ...\n",
      "   [5.11798859e-02 3.60757709e-02 9.94324684e-03]\n",
      "   [3.72302532e-01 4.34585452e-01 1.13665819e-01]\n",
      "   [5.73582053e-02 1.05097771e-01 1.65950060e-02]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[5.64520121e-01 5.22324860e-01 9.00609732e-01]\n",
      "   [4.59924817e-01 6.31840706e-01 8.00414741e-01]\n",
      "   [1.41102254e-01 3.88548553e-01 2.80000627e-01]\n",
      "   ...\n",
      "   [7.19415069e-01 5.43749392e-01 5.46945393e-01]\n",
      "   [7.86369383e-01 8.75930905e-01 8.94606590e-01]\n",
      "   [7.70979762e-01 7.17278838e-01 8.39580417e-01]]\n",
      "\n",
      "  [[5.11621356e-01 6.67049408e-01 8.74638319e-01]\n",
      "   [1.90561652e-01 3.77170920e-01 4.31503296e-01]\n",
      "   [1.51956797e-01 4.93366361e-01 3.13069046e-01]\n",
      "   ...\n",
      "   [8.00668180e-01 6.40079200e-01 7.33078122e-01]\n",
      "   [2.32444465e-01 5.19348741e-01 5.74654460e-01]\n",
      "   [4.19814318e-01 6.41923308e-01 5.27405977e-01]]\n",
      "\n",
      "  [[2.99003780e-01 3.29459935e-01 7.12441325e-01]\n",
      "   [3.76198292e-02 9.99150872e-02 2.12395787e-01]\n",
      "   [6.41851425e-02 2.43800402e-01 2.04590917e-01]\n",
      "   ...\n",
      "   [9.57155526e-02 7.87484944e-02 7.75691569e-02]\n",
      "   [8.93789530e-03 3.46463919e-02 8.39516521e-03]\n",
      "   [6.09861314e-02 1.91936910e-01 6.77466989e-02]]]]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[[[0.3585397  0.38427362 0.1873034 ]\n",
      "   [0.34275386 0.33972532 0.19669896]\n",
      "   [0.3346174  0.36325538 0.23110712]\n",
      "   ...\n",
      "   [0.18620217 0.22153682 0.20052606]\n",
      "   [0.17090386 0.2060996  0.15411419]\n",
      "   [0.23523706 0.3077317  0.22811192]]\n",
      "\n",
      "  [[0.26265436 0.3734489  0.1180574 ]\n",
      "   [0.44007817 0.45144156 0.22402799]\n",
      "   [0.3605659  0.43314347 0.22910559]\n",
      "   ...\n",
      "   [0.26853243 0.3014543  0.20022264]\n",
      "   [0.32813105 0.38888845 0.23588678]\n",
      "   [0.31875855 0.3365627  0.24535578]]\n",
      "\n",
      "  [[0.22348082 0.35225886 0.10507321]\n",
      "   [0.52097887 0.5338098  0.25171256]\n",
      "   [0.33868557 0.4970353  0.26782006]\n",
      "   ...\n",
      "   [0.30584162 0.27914783 0.20185652]\n",
      "   [0.24435437 0.32073635 0.19436553]\n",
      "   [0.32704377 0.34227562 0.23630148]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[0.39334625 0.44634235 0.50365835]\n",
      "   [0.36461502 0.52076864 0.4286363 ]\n",
      "   [0.37982216 0.47177076 0.37933496]\n",
      "   ...\n",
      "   [0.55298513 0.49980122 0.37816292]\n",
      "   [0.41513255 0.48704597 0.3941809 ]\n",
      "   [0.36045253 0.3570283  0.28015983]]\n",
      "\n",
      "  [[0.40364838 0.49321097 0.48455137]\n",
      "   [0.3209133  0.36762738 0.27308124]\n",
      "   [0.29401964 0.42732573 0.2978705 ]\n",
      "   ...\n",
      "   [0.52230155 0.44582736 0.41881222]\n",
      "   [0.3759313  0.46539503 0.4314346 ]\n",
      "   [0.34410357 0.33983257 0.28420013]]\n",
      "\n",
      "  [[0.3036638  0.2983951  0.3847881 ]\n",
      "   [0.19329783 0.2608449  0.29576987]\n",
      "   [0.31027877 0.3787033  0.2877208 ]\n",
      "   ...\n",
      "   [0.2468648  0.2133131  0.16919851]\n",
      "   [0.2800201  0.2804985  0.20595682]\n",
      "   [0.17179167 0.2251445  0.1422216 ]]]\n",
      "\n",
      "\n",
      " [[[0.32774994 0.34028098 0.14699906]\n",
      "   [0.2872279  0.24507242 0.12302792]\n",
      "   [0.31522685 0.3013755  0.13052541]\n",
      "   ...\n",
      "   [0.24892631 0.25385246 0.48897964]\n",
      "   [0.35142148 0.338552   0.5219242 ]\n",
      "   [0.43275163 0.46332443 0.6119479 ]]\n",
      "\n",
      "  [[0.2908102  0.3700318  0.12214974]\n",
      "   [0.27869457 0.2681591  0.1078853 ]\n",
      "   [0.1853804  0.28928226 0.09221277]\n",
      "   ...\n",
      "   [0.3037232  0.2984308  0.4353911 ]\n",
      "   [0.368779   0.34961808 0.48713562]\n",
      "   [0.43492183 0.3788473  0.55693597]]\n",
      "\n",
      "  [[0.19232714 0.2661323  0.0915373 ]\n",
      "   [0.16456977 0.25072247 0.08734483]\n",
      "   [0.05926827 0.18605971 0.04383183]\n",
      "   ...\n",
      "   [0.44826698 0.33728468 0.34788138]\n",
      "   [0.34152687 0.39396578 0.30217075]\n",
      "   [0.41838995 0.36943728 0.32757828]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[0.38365197 0.4208723  0.6744391 ]\n",
      "   [0.46774876 0.6221671  0.52918935]\n",
      "   [0.5145034  0.57173026 0.33897895]\n",
      "   ...\n",
      "   [0.80358344 0.60948974 0.61782944]\n",
      "   [0.7249478  0.6733411  0.7249514 ]\n",
      "   [0.72553486 0.54389393 0.673718  ]]\n",
      "\n",
      "  [[0.38735104 0.57202685 0.6572784 ]\n",
      "   [0.39226928 0.4617501  0.39984843]\n",
      "   [0.43064216 0.54516333 0.3331191 ]\n",
      "   ...\n",
      "   [0.7488644  0.5367153  0.674232  ]\n",
      "   [0.79009926 0.67973393 0.788099  ]\n",
      "   [0.60117465 0.42901853 0.5338922 ]]\n",
      "\n",
      "  [[0.36171567 0.39076513 0.5830257 ]\n",
      "   [0.22018069 0.323563   0.42495194]\n",
      "   [0.31791878 0.45706898 0.35868314]\n",
      "   ...\n",
      "   [0.18156916 0.10106003 0.21195996]\n",
      "   [0.41191912 0.35679033 0.36501276]\n",
      "   [0.5297281  0.43144006 0.44681984]]]\n",
      "\n",
      "\n",
      " [[[0.20359567 0.33798903 0.08651835]\n",
      "   [0.18258888 0.23321396 0.06786314]\n",
      "   [0.21049681 0.24693847 0.07755196]\n",
      "   ...\n",
      "   [0.1571692  0.30931738 0.06558517]\n",
      "   [0.20407793 0.3089254  0.06328338]\n",
      "   [0.27016538 0.43118292 0.14834231]]\n",
      "\n",
      "  [[0.16112542 0.32117394 0.04750645]\n",
      "   [0.26460528 0.32578182 0.08537722]\n",
      "   [0.23653859 0.35136914 0.10268089]\n",
      "   ...\n",
      "   [0.2339015  0.31227434 0.05378518]\n",
      "   [0.30492038 0.39142907 0.05307147]\n",
      "   [0.33466858 0.43116373 0.08417869]]\n",
      "\n",
      "  [[0.10556775 0.25098968 0.03227732]\n",
      "   [0.18427259 0.37267655 0.06876457]\n",
      "   [0.09967759 0.3034894  0.08823359]\n",
      "   ...\n",
      "   [0.34055364 0.3702718  0.10179672]\n",
      "   [0.22092402 0.4010965  0.05606145]\n",
      "   [0.35206556 0.49751467 0.08324832]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[0.21524477 0.30379832 0.32641035]\n",
      "   [0.18135762 0.36420023 0.1677255 ]\n",
      "   [0.4456855  0.5568708  0.27920502]\n",
      "   ...\n",
      "   [0.62160236 0.6018296  0.22114408]\n",
      "   [0.46426046 0.624321   0.23209572]\n",
      "   [0.3782043  0.45330986 0.16123423]]\n",
      "\n",
      "  [[0.12111273 0.27749413 0.16702262]\n",
      "   [0.13140303 0.28323734 0.08493012]\n",
      "   [0.28640252 0.473519   0.21768373]\n",
      "   ...\n",
      "   [0.4243228  0.54248315 0.15221283]\n",
      "   [0.36294585 0.5610304  0.11899257]\n",
      "   [0.20882106 0.43148068 0.08325264]]\n",
      "\n",
      "  [[0.14139855 0.17261606 0.21540815]\n",
      "   [0.13172984 0.251245   0.21473104]\n",
      "   [0.20154041 0.40103272 0.21009886]\n",
      "   ...\n",
      "   [0.05193484 0.11705899 0.02290916]\n",
      "   [0.09913185 0.2325775  0.03082275]\n",
      "   [0.19652477 0.46373326 0.1035324 ]]]\n",
      "\n",
      "\n",
      " ...\n",
      "\n",
      "\n",
      " [[[0.7318839  0.64816386 0.599524  ]\n",
      "   [0.7956127  0.6985676  0.6965805 ]\n",
      "   [0.8045162  0.8076983  0.8156836 ]\n",
      "   ...\n",
      "   [0.8968878  0.86110103 0.9131566 ]\n",
      "   [0.8234737  0.8249058  0.8652479 ]\n",
      "   [0.86874676 0.8651137  0.8832599 ]]\n",
      "\n",
      "  [[0.6621626  0.68143046 0.5575575 ]\n",
      "   [0.85486966 0.8228171  0.8193466 ]\n",
      "   [0.82204694 0.86748135 0.87570953]\n",
      "   ...\n",
      "   [0.9176841  0.92319405 0.9582605 ]\n",
      "   [0.87856936 0.90188396 0.9222798 ]\n",
      "   [0.8835467  0.87008774 0.92081976]]\n",
      "\n",
      "  [[0.72173935 0.7251564  0.67444026]\n",
      "   [0.9229059  0.8927529  0.83759135]\n",
      "   [0.7962307  0.8969274  0.84714425]\n",
      "   ...\n",
      "   [0.9654125  0.9257293  0.9486585 ]\n",
      "   [0.9061619  0.93537563 0.929591  ]\n",
      "   [0.9221524  0.8666672  0.8955729 ]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[0.52585584 0.5228935  0.5038367 ]\n",
      "   [0.30547756 0.36347315 0.25860077]\n",
      "   [0.42512387 0.5509356  0.41511112]\n",
      "   ...\n",
      "   [0.7673925  0.686328   0.69854224]\n",
      "   [0.69445014 0.6679284  0.69542646]\n",
      "   [0.66655225 0.64817786 0.672075  ]]\n",
      "\n",
      "  [[0.4057261  0.44750404 0.35195976]\n",
      "   [0.23029682 0.3181346  0.16841984]\n",
      "   [0.33622828 0.4302137  0.26237237]\n",
      "   ...\n",
      "   [0.7643666  0.6836211  0.70572436]\n",
      "   [0.7080838  0.71870863 0.7363124 ]\n",
      "   [0.6198868  0.5887005  0.6485178 ]]\n",
      "\n",
      "  [[0.4014857  0.3956356  0.33964846]\n",
      "   [0.3318997  0.38843715 0.3353237 ]\n",
      "   [0.3244169  0.43316185 0.2665128 ]\n",
      "   ...\n",
      "   [0.54825574 0.50468117 0.5300129 ]\n",
      "   [0.65143806 0.67103374 0.674381  ]\n",
      "   [0.6594473  0.64282984 0.63682497]]]\n",
      "\n",
      "\n",
      " [[[0.37631124 0.35827503 0.19916952]\n",
      "   [0.41825414 0.34894705 0.24238777]\n",
      "   [0.4220919  0.39374024 0.29571128]\n",
      "   ...\n",
      "   [0.26597625 0.23235595 0.41880327]\n",
      "   [0.2640803  0.20974737 0.3016691 ]\n",
      "   [0.32282907 0.31481707 0.43573064]]\n",
      "\n",
      "  [[0.32715416 0.39790222 0.15920767]\n",
      "   [0.42791677 0.40072304 0.23041084]\n",
      "   [0.35345542 0.42965713 0.2539495 ]\n",
      "   ...\n",
      "   [0.2833646  0.26895916 0.34799173]\n",
      "   [0.36040807 0.27682793 0.3908129 ]\n",
      "   [0.4120076  0.2808465  0.43575045]]\n",
      "\n",
      "  [[0.26576686 0.33059472 0.1349808 ]\n",
      "   [0.3784787  0.37507182 0.17561722]\n",
      "   [0.19594675 0.38643607 0.16003793]\n",
      "   ...\n",
      "   [0.3372267  0.19418317 0.23463595]\n",
      "   [0.27752334 0.24432051 0.19620013]\n",
      "   [0.36121005 0.22561747 0.28431183]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[0.21435758 0.25741285 0.4669734 ]\n",
      "   [0.12368754 0.22526503 0.1910711 ]\n",
      "   [0.33638018 0.3634833  0.3002542 ]\n",
      "   ...\n",
      "   [0.6299127  0.50323355 0.35936692]\n",
      "   [0.5940926  0.5298462  0.43938187]\n",
      "   [0.56958705 0.45805517 0.38803917]]\n",
      "\n",
      "  [[0.18531355 0.3424487  0.39225572]\n",
      "   [0.08603448 0.14196065 0.11788768]\n",
      "   [0.26718336 0.3366899  0.2582831 ]\n",
      "   ...\n",
      "   [0.41469735 0.36675042 0.2105462 ]\n",
      "   [0.54875946 0.56897676 0.39925438]\n",
      "   [0.44211936 0.4114431  0.28255025]]\n",
      "\n",
      "  [[0.18957216 0.19356585 0.37847754]\n",
      "   [0.07394207 0.11980045 0.16524637]\n",
      "   [0.11318317 0.16749519 0.11347768]\n",
      "   ...\n",
      "   [0.08587778 0.07358876 0.03469747]\n",
      "   [0.1012603  0.127375   0.04207629]\n",
      "   [0.20888126 0.24074966 0.10010302]]]\n",
      "\n",
      "\n",
      " [[[0.5341592  0.51270425 0.24624455]\n",
      "   [0.573915   0.49763    0.28165263]\n",
      "   [0.6397532  0.6495778  0.40459538]\n",
      "   ...\n",
      "   [0.6646457  0.6258653  0.53531295]\n",
      "   [0.62834114 0.5577213  0.48957103]\n",
      "   [0.6357129  0.6537792  0.5635654 ]]\n",
      "\n",
      "  [[0.47234824 0.5641566  0.17509645]\n",
      "   [0.6590333  0.64239687 0.33401155]\n",
      "   [0.6604981  0.73889494 0.46888822]\n",
      "   ...\n",
      "   [0.6143939  0.6383471  0.49102694]\n",
      "   [0.5874565  0.583118   0.45827085]\n",
      "   [0.62384176 0.6105192  0.50366396]]\n",
      "\n",
      "  [[0.4452505  0.5533475  0.15261441]\n",
      "   [0.71188843 0.7494435  0.33350015]\n",
      "   [0.54755384 0.75199497 0.42840227]\n",
      "   ...\n",
      "   [0.71713233 0.6666435  0.4672985 ]\n",
      "   [0.5511936  0.678306   0.3662503 ]\n",
      "   [0.6358377  0.6251838  0.44360936]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[0.28953487 0.33719975 0.21582842]\n",
      "   [0.31108493 0.4999574  0.19565529]\n",
      "   [0.46958655 0.6343643  0.24184078]\n",
      "   ...\n",
      "   [0.72777724 0.7285148  0.4610559 ]\n",
      "   [0.68290424 0.7538007  0.60714304]\n",
      "   [0.7479614  0.71789205 0.62063456]]\n",
      "\n",
      "  [[0.29330105 0.37492445 0.20432365]\n",
      "   [0.24947569 0.38084933 0.12575251]\n",
      "   [0.33014095 0.5152439  0.18588203]\n",
      "   ...\n",
      "   [0.8361139  0.79988205 0.6512468 ]\n",
      "   [0.71752685 0.83688784 0.70241797]\n",
      "   [0.62831664 0.6125079  0.521759  ]]\n",
      "\n",
      "  [[0.24413317 0.25685358 0.23706776]\n",
      "   [0.16796136 0.2653446  0.21392512]\n",
      "   [0.28499937 0.46205652 0.21303731]\n",
      "   ...\n",
      "   [0.434961   0.42201623 0.31545234]\n",
      "   [0.48451954 0.52776873 0.34238863]\n",
      "   [0.345052   0.42822    0.25290295]]]]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[[[0.30134475 0.27437586 0.13375095]\n",
      "   [0.14898437 0.1559414  0.08104396]\n",
      "   [0.12324589 0.10978347 0.07841235]\n",
      "   ...\n",
      "   [0.04398251 0.0809558  0.00937006]\n",
      "   [0.07145169 0.08701584 0.0133065 ]\n",
      "   [0.05739182 0.09264696 0.01898858]]\n",
      "\n",
      "  [[0.23163056 0.2589221  0.0897609 ]\n",
      "   [0.09508914 0.10520938 0.04528236]\n",
      "   [0.14721376 0.13367385 0.08273959]\n",
      "   ...\n",
      "   [0.07972187 0.11776841 0.00665423]\n",
      "   [0.10405567 0.15992436 0.00928143]\n",
      "   [0.05880016 0.07231033 0.00718322]]\n",
      "\n",
      "  [[0.10966811 0.19362104 0.0561409 ]\n",
      "   [0.08542791 0.13087481 0.0439463 ]\n",
      "   [0.13138878 0.17682004 0.07235396]\n",
      "   ...\n",
      "   [0.10497922 0.15200603 0.01029798]\n",
      "   [0.0817073  0.1288957  0.00625849]\n",
      "   [0.06503087 0.11117631 0.0075548 ]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[0.22345537 0.289669   0.19465852]\n",
      "   [0.21656585 0.25572395 0.26024753]\n",
      "   [0.21826667 0.2890537  0.34289557]\n",
      "   ...\n",
      "   [0.3208584  0.25239962 0.14935136]\n",
      "   [0.28672534 0.41523528 0.20184931]\n",
      "   [0.32430905 0.44165045 0.24752766]]\n",
      "\n",
      "  [[0.16409251 0.22186649 0.11697018]\n",
      "   [0.25764304 0.39648134 0.26436156]\n",
      "   [0.5221959  0.5977974  0.60786974]\n",
      "   ...\n",
      "   [0.38670453 0.33438745 0.19844604]\n",
      "   [0.15115392 0.33480167 0.16206759]\n",
      "   [0.21130782 0.34148675 0.12534651]]\n",
      "\n",
      "  [[0.17568785 0.20185664 0.1460805 ]\n",
      "   [0.27373496 0.27867746 0.26997656]\n",
      "   [0.36470068 0.4518111  0.29850075]\n",
      "   ...\n",
      "   [0.22636053 0.24272525 0.09378594]\n",
      "   [0.07587385 0.15839693 0.0256547 ]\n",
      "   [0.02202666 0.04459447 0.0103921 ]]]\n",
      "\n",
      "\n",
      " [[[0.13851672 0.15282825 0.06055608]\n",
      "   [0.04758587 0.06550324 0.02894557]\n",
      "   [0.04404676 0.0384286  0.0340023 ]\n",
      "   ...\n",
      "   [0.04657391 0.11050254 0.03763941]\n",
      "   [0.04979905 0.09394747 0.05778861]\n",
      "   [0.07940304 0.13136148 0.07798371]]\n",
      "\n",
      "  [[0.09915555 0.13848108 0.03674564]\n",
      "   [0.0294494  0.03473136 0.01312429]\n",
      "   [0.04727313 0.03915983 0.02364123]\n",
      "   ...\n",
      "   [0.07761934 0.15644825 0.03500241]\n",
      "   [0.1062108  0.18407413 0.0589762 ]\n",
      "   [0.10279709 0.16322845 0.07063049]]\n",
      "\n",
      "  [[0.02878281 0.07873243 0.01501662]\n",
      "   [0.0204145  0.04155236 0.01154351]\n",
      "   [0.03414795 0.06364506 0.01718289]\n",
      "   ...\n",
      "   [0.06336087 0.1738339  0.03274244]\n",
      "   [0.06739321 0.16228911 0.03800923]\n",
      "   [0.09511822 0.19236755 0.06824034]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[0.04126096 0.10038939 0.0587081 ]\n",
      "   [0.0178822  0.04578936 0.03155306]\n",
      "   [0.04622084 0.09037471 0.10165244]\n",
      "   ...\n",
      "   [0.09877467 0.20056495 0.07312036]\n",
      "   [0.10627437 0.25980207 0.11433518]\n",
      "   [0.10587105 0.20442608 0.05825177]]\n",
      "\n",
      "  [[0.0431684  0.0921908  0.05836031]\n",
      "   [0.01897296 0.06507972 0.03793538]\n",
      "   [0.03593314 0.08705765 0.070997  ]\n",
      "   ...\n",
      "   [0.10113347 0.2527537  0.09889519]\n",
      "   [0.05240795 0.22045302 0.06243512]\n",
      "   [0.04411602 0.13089645 0.02753848]]\n",
      "\n",
      "  [[0.08799198 0.12482625 0.09592697]\n",
      "   [0.01672462 0.03791928 0.02669728]\n",
      "   [0.0162181  0.03827769 0.01960012]\n",
      "   ...\n",
      "   [0.07118055 0.20877504 0.04713964]\n",
      "   [0.0304268  0.09167141 0.01401037]\n",
      "   [0.02019092 0.06201211 0.01512089]]]\n",
      "\n",
      "\n",
      " [[[0.55540854 0.5370184  0.29751277]\n",
      "   [0.35446593 0.35105556 0.20427203]\n",
      "   [0.33767152 0.36585438 0.23170075]\n",
      "   ...\n",
      "   [0.30038044 0.40422675 0.06059545]\n",
      "   [0.3626481  0.40575755 0.06486011]\n",
      "   [0.3399728  0.42109692 0.11151332]]\n",
      "\n",
      "  [[0.50981164 0.5469374  0.23360056]\n",
      "   [0.32244986 0.34327295 0.1921927 ]\n",
      "   [0.38688138 0.40271667 0.31625628]\n",
      "   ...\n",
      "   [0.20139551 0.31195253 0.02587688]\n",
      "   [0.25908238 0.33230576 0.04025877]\n",
      "   [0.31581154 0.36590433 0.07573152]]\n",
      "\n",
      "  [[0.3523782  0.45230833 0.14718807]\n",
      "   [0.36601055 0.43164298 0.19369102]\n",
      "   [0.28651702 0.38944137 0.20704433]\n",
      "   ...\n",
      "   [0.22078502 0.29625112 0.03340694]\n",
      "   [0.17461002 0.28338712 0.03406483]\n",
      "   [0.26599842 0.32894796 0.07757902]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[0.70148736 0.66821796 0.51852494]\n",
      "   [0.5562871  0.5513685  0.5238798 ]\n",
      "   [0.3722886  0.4502685  0.5088717 ]\n",
      "   ...\n",
      "   [0.37834898 0.35960463 0.22551978]\n",
      "   [0.4065687  0.3792929  0.29419953]\n",
      "   [0.34461114 0.33344197 0.21948135]]\n",
      "\n",
      "  [[0.68296933 0.64478004 0.37421504]\n",
      "   [0.46609437 0.5680046  0.2717089 ]\n",
      "   [0.66340125 0.7621394  0.5998308 ]\n",
      "   ...\n",
      "   [0.43507394 0.4009242  0.2568753 ]\n",
      "   [0.3318509  0.3982146  0.2024827 ]\n",
      "   [0.26949185 0.25631487 0.14390284]]\n",
      "\n",
      "  [[0.5248804  0.4822198  0.32711044]\n",
      "   [0.53484315 0.51577127 0.33711475]\n",
      "   [0.544492   0.59344953 0.3554212 ]\n",
      "   ...\n",
      "   [0.2698332  0.34668726 0.1048283 ]\n",
      "   [0.13094026 0.18561596 0.04087445]\n",
      "   [0.11014593 0.1475411  0.05226496]]]\n",
      "\n",
      "\n",
      " ...\n",
      "\n",
      "\n",
      " [[[0.31212962 0.4404638  0.29327017]\n",
      "   [0.17971972 0.27187258 0.22305962]\n",
      "   [0.15927023 0.250331   0.2724157 ]\n",
      "   ...\n",
      "   [0.2584526  0.4503223  0.4699983 ]\n",
      "   [0.2376743  0.41549236 0.48083356]\n",
      "   [0.29036394 0.48345295 0.4642437 ]]\n",
      "\n",
      "  [[0.16961277 0.36611253 0.21109971]\n",
      "   [0.17803285 0.2825096  0.2709935 ]\n",
      "   [0.24011093 0.30830914 0.3826643 ]\n",
      "   ...\n",
      "   [0.23312327 0.37074378 0.42755318]\n",
      "   [0.31320515 0.490287   0.50995   ]\n",
      "   [0.33294612 0.51033026 0.5687433 ]]\n",
      "\n",
      "  [[0.09187749 0.29776245 0.14658958]\n",
      "   [0.17397818 0.36990443 0.31410405]\n",
      "   [0.13197133 0.3388957  0.26223725]\n",
      "   ...\n",
      "   [0.33824202 0.47309333 0.42773288]\n",
      "   [0.2506343  0.47303793 0.47309694]\n",
      "   [0.33036143 0.51329327 0.53500456]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[0.08116502 0.2940016  0.23531675]\n",
      "   [0.18830413 0.44895604 0.38105997]\n",
      "   [0.17589355 0.43694335 0.4253495 ]\n",
      "   ...\n",
      "   [0.15566683 0.38127127 0.42852986]\n",
      "   [0.11002693 0.42897034 0.5036918 ]\n",
      "   [0.11215985 0.3613367  0.39422476]]\n",
      "\n",
      "  [[0.08884203 0.32057494 0.21186453]\n",
      "   [0.15841952 0.39745742 0.33652696]\n",
      "   [0.19774091 0.45870608 0.41542864]\n",
      "   ...\n",
      "   [0.16435635 0.42032582 0.4956461 ]\n",
      "   [0.05327123 0.4128001  0.37424475]\n",
      "   [0.05163693 0.28920943 0.25636744]]\n",
      "\n",
      "  [[0.18437928 0.342587   0.31546718]\n",
      "   [0.09744421 0.2373364  0.2285189 ]\n",
      "   [0.10222048 0.27400482 0.25536138]\n",
      "   ...\n",
      "   [0.10668647 0.34185734 0.22572583]\n",
      "   [0.06020451 0.29210854 0.14144388]\n",
      "   [0.04718798 0.25063384 0.17304027]]]\n",
      "\n",
      "\n",
      " [[[0.22426003 0.19698375 0.07107145]\n",
      "   [0.13882172 0.10106915 0.03439128]\n",
      "   [0.15683487 0.041657   0.0133177 ]\n",
      "   ...\n",
      "   [0.09934768 0.18001679 0.02789482]\n",
      "   [0.12749374 0.18986213 0.04116869]\n",
      "   [0.10180131 0.16506118 0.04025865]]\n",
      "\n",
      "  [[0.19848984 0.21428603 0.04210064]\n",
      "   [0.07874951 0.04914665 0.0084537 ]\n",
      "   [0.0526264  0.02602178 0.00167787]\n",
      "   ...\n",
      "   [0.1363172  0.21525496 0.01878425]\n",
      "   [0.16974148 0.2720164  0.03202063]\n",
      "   [0.14940733 0.19698456 0.03375787]]\n",
      "\n",
      "  [[0.13259965 0.19178343 0.04405016]\n",
      "   [0.07613328 0.09543988 0.01935962]\n",
      "   [0.09316194 0.08061293 0.01402685]\n",
      "   ...\n",
      "   [0.12737876 0.2451365  0.0233922 ]\n",
      "   [0.11677092 0.20119831 0.02472907]\n",
      "   [0.1469416  0.24562421 0.04115695]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[0.20352054 0.21357584 0.11235365]\n",
      "   [0.16909778 0.19966602 0.12041929]\n",
      "   [0.22039923 0.3166117  0.24580619]\n",
      "   ...\n",
      "   [0.26827228 0.24791768 0.17861307]\n",
      "   [0.2535153  0.25847548 0.18576503]\n",
      "   [0.2424432  0.23234153 0.13180235]]\n",
      "\n",
      "  [[0.19836193 0.2027936  0.09207702]\n",
      "   [0.17603058 0.23875958 0.09167296]\n",
      "   [0.2328527  0.4099627  0.24323446]\n",
      "   ...\n",
      "   [0.30841565 0.25007564 0.23607719]\n",
      "   [0.24020413 0.3099081  0.22327822]\n",
      "   [0.2272594  0.24955228 0.13185534]]\n",
      "\n",
      "  [[0.16822553 0.16374698 0.07988214]\n",
      "   [0.17306358 0.19933954 0.08695504]\n",
      "   [0.26193976 0.35448748 0.16218531]\n",
      "   ...\n",
      "   [0.19414005 0.20247865 0.11846873]\n",
      "   [0.15364662 0.19515872 0.07425615]\n",
      "   [0.08610711 0.11130106 0.05300578]]]\n",
      "\n",
      "\n",
      " [[[0.32443202 0.41282934 0.24069315]\n",
      "   [0.19754285 0.27087596 0.19585037]\n",
      "   [0.17652369 0.2492241  0.2274862 ]\n",
      "   ...\n",
      "   [0.0885008  0.24031997 0.14557499]\n",
      "   [0.08006883 0.19996554 0.20323181]\n",
      "   [0.12391847 0.23591048 0.19036159]]\n",
      "\n",
      "  [[0.19491065 0.33991164 0.15691125]\n",
      "   [0.17494199 0.25239944 0.19085175]\n",
      "   [0.2302309  0.27394015 0.29116458]\n",
      "   ...\n",
      "   [0.14653975 0.32507333 0.20590115]\n",
      "   [0.18129122 0.353151   0.266607  ]\n",
      "   [0.17838213 0.31103975 0.25271896]]\n",
      "\n",
      "  [[0.1108571  0.2995121  0.10888049]\n",
      "   [0.16442874 0.3118871  0.20180237]\n",
      "   [0.13493532 0.29145145 0.19556403]\n",
      "   ...\n",
      "   [0.14416605 0.38075885 0.20737323]\n",
      "   [0.12245795 0.34825003 0.21597922]\n",
      "   [0.17807138 0.38088998 0.31613982]]\n",
      "\n",
      "  ...\n",
      "\n",
      "  [[0.18200624 0.3874643  0.31427374]\n",
      "   [0.3429222  0.5375554  0.49865547]\n",
      "   [0.23626277 0.4218767  0.4701955 ]\n",
      "   ...\n",
      "   [0.17723191 0.3412008  0.18024391]\n",
      "   [0.16315293 0.41907886 0.21239597]\n",
      "   [0.18097878 0.40877506 0.1835025 ]]\n",
      "\n",
      "  [[0.15411356 0.34722257 0.24211597]\n",
      "   [0.30696368 0.52120125 0.4359846 ]\n",
      "   [0.22577557 0.4801916  0.35296997]\n",
      "   ...\n",
      "   [0.18815458 0.40204626 0.20911235]\n",
      "   [0.11399093 0.45462412 0.14629096]\n",
      "   [0.09347814 0.33896157 0.1016891 ]]\n",
      "\n",
      "  [[0.20127675 0.34475833 0.28524992]\n",
      "   [0.2238096  0.39817193 0.3312025 ]\n",
      "   [0.21551102 0.38810724 0.31721097]\n",
      "   ...\n",
      "   [0.14816058 0.34691405 0.10819244]\n",
      "   [0.08702356 0.26232687 0.05676574]\n",
      "   [0.06380665 0.2644951  0.07931697]]]]\n"
     ]
    },
    {
     "ename": "KeyError",
     "evalue": "3",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-10-c5993b8d3d10>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m \u001b[0mgenerate_images\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;32m<ipython-input-8-70540361faa8>\u001b[0m in \u001b[0;36mgenerate_images\u001b[1;34m()\u001b[0m\n\u001b[0;32m     53\u001b[0m     \u001b[1;32mfor\u001b[0m \u001b[0mcn\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mrange\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mlen\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mcaption_vectors\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m-\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     54\u001b[0m         \u001b[0mcaption_images\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;33m[\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 55\u001b[1;33m         \u001b[1;32mfor\u001b[0m \u001b[0mi\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mim\u001b[0m \u001b[1;32min\u001b[0m \u001b[0menumerate\u001b[0m\u001b[1;33m(\u001b[0m \u001b[0mcaption_image_dic\u001b[0m\u001b[1;33m[\u001b[0m \u001b[0mcn\u001b[0m \u001b[1;33m]\u001b[0m \u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     56\u001b[0m             \u001b[1;31m# im_name = \"caption_{}_{}.jpg\".format(cn, i)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     57\u001b[0m             \u001b[1;31m# scipy.misc.imsave( join(args.data_dir, 'val_samples/{}'.format(im_name)) , im)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mKeyError\u001b[0m: 3"
     ]
    }
   ],
   "source": [
    "generate_images()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def make_video(images, fps=25, size=None,\n",
    "               is_color=True, format=\"XVID\"):\n",
    "    \"\"\"\n",
    "    Create a video from a list of images.\n",
    " \n",
    "    @param      outvid      output video\n",
    "    @param      images      list of images to use in the video\n",
    "    @param      fps         frame per second\n",
    "    @param      size        size of each frame\n",
    "    @param      is_color    color\n",
    "    @param      format      see http://www.fourcc.org/codecs.php\n",
    "    @return                 see http://opencv-python-tutroals.readthedocs.org/en/latest/py_tutorials/py_gui/py_video_display/py_video_display.html\n",
    " \n",
    "    The function relies on http://opencv-python-tutroals.readthedocs.org/en/latest/.\n",
    "    By default, the video will have the size of the first image.\n",
    "    It will resize every image to this size before adding them to the video.\n",
    "    \"\"\"\n",
    "    from cv2 import VideoWriter, VideoWriter_fourcc, imread, resize\n",
    "    fourcc = VideoWriter_fourcc(*format)\n",
    "    vid = cv2.VideoWriter('output.avi', fourcc, 25.0, (64,64))\n",
    "    \n",
    "    for image in images:\n",
    "        img = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)\n",
    "        vid.write(img)\n",
    "    vid.release()\n",
    "    cv2.destroyAllWindows()\n",
    "    return"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.python import pywrap_tensorflow\n",
    "import os\n",
    "\n",
    "checkpoint_path = os.path.join('', 'Data/Models/latest_model_flowers_temp.ckpt')\n",
    "reader = pywrap_tensorflow.NewCheckpointReader(checkpoint_path)\n",
    "var_to_shape_map = reader.get_variable_to_shape_map()\n",
    "\n",
    "for key in var_to_shape_map:\n",
    "    print(\"tensor_name: \", key)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}