Chap13 - CNN 완료

ExcelsiorCJH · Oct 19, 2018 · a52e0b9 · a52e0b9
1 parent 856c34d
commit a52e0b9
Showing 1 changed file with 17 additions and 25 deletions.
diff --git a/Chap13-Convolutional_Neural_Networks/Chap13-Convolutional_Neural_Networks.ipynb b/Chap13-Convolutional_Neural_Networks/Chap13-Convolutional_Neural_Networks.ipynb
@@ -561,31 +561,14 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 22,
+   "execution_count": 48,
    "metadata": {},
    "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "fmap.shape : (7, 7, 1, 2)\n"
-     ]
-    },
     {
      "data": {
-      "image/png": "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\n",
+      "image/png": "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\n",
       "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
-      ]
-     },
-     "metadata": {},
-     "output_type": "display_data"
-    },
-    {
-     "data": {
-      "image/png": "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\n",
-      "text/plain": [
-       "<Figure size 432x288 with 1 Axes>"
+       "<Figure size 1008x360 with 2 Axes>"
       ]
      },
      "metadata": {},
@@ -596,11 +579,20 @@
     "fmap = np.zeros(shape=(7, 7, 1, 2), dtype=np.float32)\n",
     "fmap[:, 3, 0, 0] = 1\n",
     "fmap[3, :, 0, 1] = 1\n",
-    "print('fmap.shape :', fmap.shape)\n",
-    "plot_image(fmap[:, :, 0, 0])\n",
-    "plt.show()\n",
-    "plot_image(fmap[:, :, 0, 1])\n",
-    "plt.show()"
+    "\n",
+    "# matplotlib으로 plotting하기\n",
+    "fig, axs = plt.subplots(1, 2, figsize=(14, 5))\n",
+    "pl_images = [\n",
+    "    fmap[:, :, 0, 0],  # vertical-filter\n",
+    "    fmap[:, :, 0, 1],  # horizontal-filter\n",
+    "]\n",
+    "\n",
+    "titles = ['vertical-filter', 'horizontal-filter']\n",
+    "\n",
+    "for i, (image, title) in enumerate(zip(pl_images, titles)):\n",
+    "    axs[i].imshow(image, cmap='gray')\n",
+    "    axs[i].set_title(title)\n",
+    "    axs[i].axis('off')"
    ]
   },
   {