{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Copyright 2020 Erik Härkönen. All rights reserved.\n",
    "# This file is licensed to you under the Apache License, Version 2.0 (the \"License\");\n",
    "# you may not use this file except in compliance with the License. You may obtain a copy\n",
    "# of the License at http://www.apache.org/licenses/LICENSE-2.0\n",
    "\n",
    "# Unless required by applicable law or agreed to in writing, software distributed under\n",
    "# the License is distributed on an \"AS IS\" BASIS, WITHOUT WARRANTIES OR REPRESENTATIONS\n",
    "# OF ANY KIND, either express or implied. See the License for the specific language\n",
    "# governing permissions and limitations under the License.\n",
    "\n",
    "%matplotlib inline\n",
    "from notebook_init import *\n",
    "\n",
    "out_root = Path('out/figures/biggan_style')\n",
    "makedirs(out_root, exist_ok=True)\n",
    "\n",
    "model = get_model('BigGAN-512', 'husky', device)\n",
    "rand = lambda : np.random.randint(np.iinfo(np.int32).max)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "classes = ['husky', 'church']\n",
    "base_seeds = [665823877, 419650361]\n",
    "style_seeds = [1922988331, 286873059, 1376693511, 1853453896]\n",
    "print(base_seeds, style_seeds)\n",
    "\n",
    "num_keep = [1, 4, 8] # switch latent after first, fourth, and eighth layer\n",
    "layer_names = ['layer1', 'layer4', 'layer8']\n",
    "\n",
    "model.truncation = 0.9\n",
    "\n",
    "for class_idx, class_name in enumerate(classes):\n",
    "    print(class_name, base_seeds[class_idx])\n",
    "    \n",
    "    model.set_output_class(class_name)\n",
    "    \n",
    "    for n_base in num_keep:\n",
    "        strip = []\n",
    "    \n",
    "        # Base\n",
    "        z0 = model.sample_latent(1, seed=base_seeds[class_idx])\n",
    "        out = model.sample_np(z0)\n",
    "        \n",
    "        # Resample style\n",
    "        plt.figure(figsize=(25,25))\n",
    "        for img_idx, seed in enumerate(style_seeds):\n",
    "            z1 = model.sample_latent(1, seed=seed)\n",
    "            \n",
    "            # Use style latent after 'n_base' layers\n",
    "            n_style = model.get_max_latents() - n_base\n",
    "            z = [z0] * n_base + [z1] * n_style\n",
    "            \n",
    "            img = model.sample_np(z)\n",
    "            strip.append(img)\n",
    "            \n",
    "            # Save individually\n",
    "            layer_name = f'layer{n_base}'\n",
    "            img_name = out_root / f'style_resample_{class_name}_{layer_name}_{img_idx}.png'\n",
    "            im = Image.fromarray((255*img).astype(np.uint8))\n",
    "            im.save(img_name)\n",
    "            \n",
    "        # Show strip\n",
    "        plt.imshow(np.hstack(strip))\n",
    "        plt.axis('off')\n",
    "        plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.7.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}