Explore The Activations.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "sys.path.insert(0, '../lucid')\n",
    "\n",
    "import os\n",
    "os.environ['CUDA_VISIBLE_DEVICES'] = '-1'\n",
    "\n",
    "import lucid_video as L\n",
    "import common as C\n",
    "import tensorflow as tf\n",
    "\n",
    "import lucid.optvis.render as render\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "11d73838ca2f49c9885805bc8dfd97dc",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=13), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e98b3f9edf024369a3618ffe0e9af751",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=19), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "def compute_c3d_acts(vid):\n",
    "    model = L.I3D()\n",
    "    model.load_graphdef()\n",
    "\n",
    "    layers = [ l.name for l in model.layers ] \n",
    "\n",
    "    acts = {}\n",
    "\n",
    "    with tf.Graph().as_default(), tf.Session():\n",
    "        t_input = tf.placeholder_with_default(tf.to_float(vid), [None, None, None, 3])\n",
    "        T = render.import_model(model, t_input, t_input)\n",
    "        for layer in L.tqdm(layers):\n",
    "            k = layer.split('/')[1].lower().replace('_', '')\n",
    "            acts[k] = T(layer).eval()\n",
    "\n",
    "    return acts\n",
    "\n",
    "def compute_c2d_acts(img): \n",
    "    model = models.InceptionV1()\n",
    "    model.load_graphdef()\n",
    "\n",
    "    layers = [ l.name for l in model.layers ] \n",
    "    acts = {}\n",
    "\n",
    "    with tf.Graph().as_default(), tf.Session():\n",
    "        t_input =  tf.placeholder_with_default(tf.to_float(img), [None, None, 3])\n",
    "        T = render.import_model(model, t_input, t_input)\n",
    "        for layer in L.tqdm(layers):\n",
    "            acts[layer] = T(layer).eval()\n",
    "    \n",
    "    return acts\n",
    "    c2d_acts = acts.copy()\n",
    "    \n",
    "vid = C.file_to_array(C.cricket_video)\n",
    "\n",
    "c3d_acts = compute_c3d_acts(vid)\n",
    "c2d_acts = compute_c2d_acts(vid[0])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f68cc54dabf74bdfb988c149b0392110",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=13), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "\n",
    "\n",
    "vid_single_frame =  np.stack([ vid[0] for _ in range(vid.shape[0])])\n",
    "\n",
    "c3d_acts_single_frame = compute_c3d_acts(vid_single_frame)\n",
    " "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7f52d79eaeb8>"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(c2d_acts['mixed4b'][0].mean(axis=-1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7f52d7a93c88>"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(c3d_acts['mixed4b'][0].mean(axis=-1).mean(0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7f52a173e550>"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(c3d_acts_single_frame['mixed4b'][0].mean(-1).mean(0))"
   ]
  }
 ],
 "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}