{ "cells": [ { "cell_type": "markdown", "metadata": { "_cell_guid": "33442feb-6868-af07-38c4-c7fb46afede2" }, "source": [ "# In this notebook, we'll show how to explore data in the TFRecord format on a smaller scale. To reach the scalability that TensorFlow/Google Cloud can do, please refer to the code in their github https://github.com/google/youtube-8m" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "0d24983c-6c8c-829d-2423-5db0cc6d5ced" }, "outputs": [], "source": [ "# This Python 3 environment comes with many helpful analytics libraries installed\n", "# It is defined by the kaggle/python docker image: https://github.com/kaggle/docker-python\n", "# For example, here's several helpful packages to load in \n", "\n", "import numpy as np # linear algebra\n", "import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n", "\n", "# Input data files are available in the \"../input/\" directory.\n", "# For example, running this (by clicking run or pressing Shift+Enter) will list the files in the input directory\n", "\n", "from subprocess import check_output\n", "print(check_output([\"ls\", \"../input\"]).decode(\"utf8\"))\n", "\n", "# Any results you write to the current directory are saved as output." ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "d27744ac-7d20-6535-1ecd-c3f81a04567b" }, "outputs": [], "source": [ "import tensorflow as tf\n", "import numpy as np\n", "from IPython.display import YouTubeVideo\n", "\n", "\n", "video_lvl_record = \"../input/video_level/train-1.tfrecord\"\n", "frame_lvl_record = \"../input/frame_level/train-1.tfrecord\"" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "26264c1a-1ed5-ce2c-44e0-a2e2296ff458" }, "outputs": [], "source": [ "vid_ids = []\n", "labels = []\n", "mean_rgb = []\n", "mean_audio = []\n", "\n", "for example in tf.python_io.tf_record_iterator(video_lvl_record):\n", " tf_example = tf.train.Example.FromString(example)\n", "\n", " vid_ids.append(tf_example.features.feature['video_id'].bytes_list.value[0].decode(encoding='UTF-8'))\n", " labels.append(tf_example.features.feature['labels'].int64_list.value)\n", " mean_rgb.append(tf_example.features.feature['mean_rgb'].float_list.value)\n", " mean_audio.append(tf_example.features.feature['mean_audio'].float_list.value)" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "a25d68b9-2e5a-470f-4f18-2c5762948049" }, "outputs": [], "source": [ "print('Number of videos in this tfrecord: ',len(mean_rgb))\n", "print('First video feature length',len(mean_rgb[0]))\n", "print('First 20 features of the first youtube video (',vid_ids[0],')')\n", "print(mean_rgb[0][:20])" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "23cafa43-2562-821f-5a84-e91131812b71" }, "outputs": [], "source": [ "def play_one_vid(record_name, video_index):\n", " return vid_ids[video_index]\n", " \n", "# this worked on my local jupyter notebook, but doesn't show on kaggle kernels:\n", "YouTubeVideo(play_one_vid(video_lvl_record, 7))" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "e4f90219-9815-2c6e-f48a-315a718bc0f7" }, "outputs": [], "source": [ "print('mean_audio has length of: ')\n", "print([len(x) for x in mean_audio][:5])\n", "print('mean_rgb has length of: ')\n", "print([len(x) for x in mean_rgb][:5])" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "2c1cd925-f7f0-3b81-4630-86ff8b1f407e" }, "outputs": [], "source": [ "# now, let's read the frame-level data\n", "# due to execution time, we're only going to read the first video\n", "\n", "feat_rgb = []\n", "feat_audio = []\n", "\n", "for example in tf.python_io.tf_record_iterator(frame_lvl_record): \n", " tf_seq_example = tf.train.SequenceExample.FromString(example)\n", " n_frames = len(tf_seq_example.feature_lists.feature_list['audio'].feature)\n", " sess = tf.InteractiveSession()\n", " rgb_frame = []\n", " audio_frame = []\n", " # iterate through frames\n", " for i in range(n_frames):\n", " rgb_frame.append(tf.cast(tf.decode_raw(\n", " tf_seq_example.feature_lists.feature_list['rgb'].feature[i].bytes_list.value[0],tf.uint8)\n", " ,tf.float32).eval())\n", " audio_frame.append(tf.cast(tf.decode_raw(\n", " tf_seq_example.feature_lists.feature_list['audio'].feature[i].bytes_list.value[0],tf.uint8)\n", " ,tf.float32).eval())\n", " \n", " \n", " sess.close()\n", " feat_rgb.append(rgb_frame)\n", " feat_audio.append(audio_frame)\n", " break" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "fa352887-c661-bb07-aa4c-c1bc2afc93c6" }, "outputs": [], "source": [ "print('The first video has %d frames' %len(feat_rgb[0]))" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "2189f322-5ab7-d928-2b6a-7f1d2945363c" }, "outputs": [], "source": [ "from matplotlib import pyplot as plt\n", "%matplotlib inline\n", "from sklearn.manifold import TSNE\n", "import numpy as np" ] }, { "cell_type": "markdown", "metadata": { "_cell_guid": "cb51a109-e999-7bee-d8e0-43ce3b1c5013" }, "source": [ "# now let's explore a little on the labels\n", "Find the most commonly appeared label in this record:" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "0fd1f8de-82ea-3fa7-de9e-28b86c197e28" }, "outputs": [], "source": [ "n=10\n", "from collections import Counter\n", "label_mapping = pd.Series.from_csv('../input/label_names.csv',header=0).to_dict()\n", "\n", "top_n = Counter([item for sublist in labels for item in sublist]).most_common(n)\n", "top_n_labels = [int(i[0]) for i in top_n]\n", "top_n_label_names = [label_mapping[x] for x in top_n_labels]\n", "top_n_label_names" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "a24caad7-a3c8-55cb-0ae8-356014e2b327" }, "outputs": [], "source": [ "import networkx as nx\n", "from itertools import combinations\n", "\n", "G=nx.Graph()\n", "\n", "G.clear()\n", "for list_of_nodes in labels:\n", " filtered_nodes = set(list_of_nodes).intersection(set(top_n_labels)) \n", " for node1,node2 in list(combinations(filtered_nodes,2)): \n", " node1_name = label_mapping[node1]\n", " node2_name = label_mapping[node2]\n", " G.add_node(node1_name)\n", " G.add_node(node2_name)\n", " G.add_edge(node1_name, node2_name)\n", "\n", "nx.draw_networkx(G)" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "01a7ef6a-4109-83cf-53f0-2446bd992515" }, "outputs": [], "source": [ "colors = plt.cm.rainbow(np.linspace(0, 1, n))\n", "mean_rgb_top_n = []\n", "labels_for_tsne = []\n", "# filtering mean_rgb so it only contains top n labels\n", "for idx, list_of_nodes in enumerate(labels):\n", " for node in list_of_nodes:\n", " if node in top_n_labels:\n", " mean_rgb_top_n.append(mean_rgb[idx])\n", " labels_for_tsne.append(node)\n", "\n", "\n", "X_embedded = TSNE(n_components=2, random_state=0).fit_transform(mean_rgb_top_n) \n", "\n", "\n", "fig = plt.figure()\n", "ax = fig.add_subplot(111)\n", "\n", "handles = []\n", "for indx, color in enumerate(colors):\n", " this_label = top_n_labels[indx]\n", " X_embedded_filtered = X_embedded[np.array([x==this_label for x in labels_for_tsne])]\n", " handles.append(ax.scatter(X_embedded_filtered[:, 0], X_embedded_filtered[:, 1], c=color, marker=\"o\",edgecolor='none'))\n", "\n", "ax.legend(handles, top_n_labels)\n", "\n", "plt.show()" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "_cell_guid": "ea548c00-8c98-5806-317b-a64552c87a84" }, "outputs": [], "source": [ "" ] } ], "metadata": { "_change_revision": 0, "_is_fork": false, "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": "3.6.0" } }, "nbformat": 4, "nbformat_minor": 0 }