Minor updates to S3D MIL-NCE colab

PiperOrigin-RevId: 316747124
margaretmz · Jun 16, 2020 · b19e1ff · b19e1ff
1 parent e3adc8e
commit b19e1ff
Showing 1 changed file with 10 additions and 7 deletions.
diff --git a/examples/colab/text_to_video_retrieval_with_s3d_milnce.ipynb b/examples/colab/text_to_video_retrieval_with_s3d_milnce.ipynb
@@ -33,7 +33,7 @@
         "id": "z2_BHI6XdJ30"
       },
       "source": [
-        "# Text-to-Video retrieval with S3D MIL-NCE using the TF-Hub module"
+        "# Text-to-Video retrieval with S3D MIL-NCE"
       ]
     },
     {
@@ -89,7 +89,11 @@
         "id": "ZxwaK-jf7qkW"
       },
       "source": [
-        "## Import TF-Hub module"
+        "## Import TF-Hub model\n",
+        "\n",
+        "This tutorial demonstrates how to use the [S3D MIL-NCE model](https://tfhub.dev/deepmind/mil-nce/s3d/1) from TensorFlow Hub to do **text-to-video retrieval** to find the most similar videos for a given text query.\n",
+        "\n",
+        "The model has 2 signatures, one for generating *video embeddings* and one for generating *text embeddings*. We will use these embedding to find the nearest neighbors in the embedding space."
       ]
     },
     {
@@ -186,8 +190,7 @@
         "    html += '\u003cimg src=\"{}\" height=\"224\"\u003e'.format(url)\n",
         "    html += '\u003c/td\u003e'\n",
         "  html += '\u003c/tr\u003e\u003c/table\u003e'\n",
-        "  return html\n",
-        "\n"
+        "  return html\n"
       ]
     },
     {
@@ -200,7 +203,7 @@
       },
       "outputs": [],
       "source": [
-        "# @title Load example images and define text queries  { display-mode: \"form\" }\n",
+        "# @title Load example videos and define text queries  { display-mode: \"form\" }\n",
         "\n",
         "video_1_url = 'https://upload.wikimedia.org/wikipedia/commons/b/b0/YosriAirTerjun.gif' # @param {type:\"string\"}\n",
         "video_2_url = 'https://upload.wikimedia.org/wikipedia/commons/e/e6/Guitar_solo_gif.gif' # @param {type:\"string\"}\n",
@@ -245,7 +248,7 @@
         "# Prepare text input.\n",
         "words_np = np.array(all_queries_video)\n",
         "\n",
-        "# Run the graph.\n",
+        "# Generate the video and text embeddings.\n",
         "video_embd, text_embd = generate_embeddings(hub_model, videos_np, words_np)\n",
         "\n",
         "# Scores between video and text is computed by dot products.\n",
@@ -274,7 +277,7 @@
   "metadata": {
     "colab": {
       "collapsed_sections": [],
-      "name": "text_to_video_retrieval_with_s3d_milnce.ipynb",
+      "name": "Text-to-Video retrieval with S3D MIL-NCE",
       "private_outputs": true,
       "provenance": [],
       "toc_visible": true