{"cells":[{"metadata":{},"cell_type":"markdown","source":"# Generative Dog Images\n\n\"https://thehappypuppysite.com/wp-content/uploads/2018/07/how-long-do-golden-retrievers-live-KH-long.jpg\"\n\n## Experiment with creating puppy pics"},{"metadata":{},"cell_type":"markdown","source":"# Table of Content\n* [General information](#General-information)\n* [Visualisation of pictures](#Visualisation-of-pictures)\n* [Breeds of dogs](#Breeds-of-dogs)\n* [Local conclusion](#Local-conclusion)"},{"metadata":{},"cell_type":"markdown","source":"# General information\n* **Task:** to generate pictures of dogs with about $20k$ different pictures of dogs given. \"This ~~person~~ ~~rental~~ ~~vessel~~ ~~waifu~~ ~~[etc.](https://thisxdoesnotexist.com/)~~ **dog** does not exist\".\n\n* **Data:** the data is [Stanford Dogs Dataset](http://vision.stanford.edu/aditya86/ImageNetDogs/) built from ImageNet. But, one of $20580$ images doesn't include, so we have $20579$.\n * 120 dog breeds, from 148 to 252 photos per breed, with 75% quantile equal to 186 photos per breed.\n * Format of pictures name: `ImageNet `[`synset`](https://en.wikipedia.org/wiki/Synonym_ring)` _ some number`. E.g. `n02090721_42` -- [Irish Wolfhound](https://en.wikipedia.org/wiki/Irish_wolfhound), picture with index 42.\n * Here is a [poster](http://vision.stanford.edu/documents/KhoslaJayadevaprakashYaoFeiFei_FGVC2011.pdf) of a dataset.\n* **Metric:** the metric is MiFID (Memorization-informed Fréchet Inception Distance). Explanation on the [forum](https://www.kaggle.com/c/generative-dog-images/discussion/97809#latest-564085) or in the [overview]((https://www.kaggle.com/c/generative-dog-images/overview/evaluation)).\n* [Kaggle overview](https://www.kaggle.com/c/generative-dog-images/overview)\n\n[To table of Content](#Table-of-Content)\n***"},{"metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true,"_kg_hide-input":true},"cell_type":"code","source":"# import\nimport os\nimport numpy as np\nimport pandas as pd\nimport matplotlib.pyplot as plt\nimport matplotlib.image as mpimg\nimport seaborn as sns\n%matplotlib inline\nplt.style.use('ggplot')\nnp.random.seed(2)\n\n\nimport keras\nprint(keras.backend.tensorflow_backend._get_available_gpus())\nprint(keras.backend.image_data_format()) # set_image_data_format('channels_first')\n\nimport warnings\nwarnings.filterwarnings('ignore')","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"# Visualisation of pictures"},{"metadata":{},"cell_type":"markdown","source":"Interesting data facts are:\n\n* There are pictures with more than one dog (even with $3$ dogs);\n* There are pictures with the dog (-s) and person (people);\n* There are pictures with more than one person (even with $4$ people);\n* There are pictures where dogs occupy less than $1/5$ of the picture;\n\n* There are pictures with text (magazine covers, from dog shows, memes and pictures with text);\n* Even wild predators included, e.g. [African wild dog](https://en.wikipedia.org/wiki/African_wild_dog) or [Dingo](https://en.wikipedia.org/wiki/Dingo), but not wolves.\n\n\nLet's visualize $9$ random pictures of a given dataset."},{"metadata":{"trusted":true,"_kg_hide-input":true},"cell_type":"code","source":"PATH = '../input/all-dogs/all-dogs/'\nPATH_LIST = os.listdir(PATH)\nprint(f'There are #{len(os.listdir(PATH))} pictures of dogs.')\n\nfig, axes = plt.subplots(nrows=3, ncols=3, figsize=(12,10))\n\nfor indx, axis in enumerate(axes.flatten()):\n rnd_indx = np.random.randint(0, len(os.listdir(PATH)))\n # https://matplotlib.org/users/image_tutorial.html\n img = plt.imread(PATH + PATH_LIST[rnd_indx])\n imgplot = axis.imshow(img)\n axis.set_title(PATH_LIST[rnd_indx])\n axis.set_axis_off()\nplt.tight_layout(rect=[0, 0.03, 1, 0.95])","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"[To table of Content](#Table-of-Content)\n***"},{"metadata":{},"cell_type":"markdown","source":"# Breeds of dogs"},{"metadata":{},"cell_type":"markdown","source":"\n\n \n \n \n \n \n \n \n \n
\n \n \"Borzoi\"\n \n\"Standard\n\n\"Great\n\"Irish\n\n\"Border\n\n\"German\n
"},{"metadata":{},"cell_type":"markdown","source":"* **Silhouettes upper are clickable**. A right-click opens the [ImageNetDogs](http://vision.stanford.edu/aditya86/ImageNetDogs/) page with pictures of this breed.\n* There are $120$ **different by size, hairiness and other body characteristics** breeds of dogs.\n* Lower you can see some visualization of photo distribution of breeds."},{"metadata":{"trusted":true,"_kg_hide-input":true},"cell_type":"code","source":"from collections import defaultdict\nby_breeds_dict = defaultdict(list)\nfor breed_code_and_pict_indx in PATH_LIST:\n breed_code, pict_indx = breed_code_and_pict_indx.split('_')\n by_breeds_dict[breed_code].append(pict_indx) ","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_kg_hide-input":true},"cell_type":"code","source":"df_aux = pd.DataFrame.from_dict(by_breeds_dict, orient='index').T\nprint(f'There are {df_aux.shape[1]} breeds')\n\nfig = plt.figure(figsize=(12, 6))\nfig.suptitle('Distribution of number of photos per breed')\nfig.add_subplot(121)\nsns.boxplot(df_aux.shape[0] - df_aux.isnull().sum())\nplt.xlabel('# of photos')\n#plt.xticks([])\nfig.add_subplot(122)\nsns.distplot(df_aux.shape[0] - df_aux.isnull().sum())\nplt.tight_layout(rect=[0, 0.03, 1, 0.95])\nplt.show()","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"[To table of Content](#Table-of-Content)\n***"},{"metadata":{},"cell_type":"markdown","source":"# Local conclusion\n**Maybe knowledge of breeds features, distribution of photos and visual analysis of pictures may help in this task of an image generation.** \n\n_Also, it is possible to get poses estimation and number of dogs per some pictures from the [ImageNetDogs](http://vision.stanford.edu/aditya86/ImageNetDogs/) page, but the question is it qualifies as External data (\"No external data can be added as a data source\" and \"C. External Data. You may use data other than the Competition Data (“External Data”) to develop and test your models and Submissions...\")_"},{"metadata":{},"cell_type":"markdown","source":"[To table of Content](#Table-of-Content)\n***"},{"metadata":{"trusted":true},"cell_type":"code","source":"","execution_count":null,"outputs":[]}],"metadata":{"kernelspec":{"display_name":"Python 3","language":"python","name":"python3"},"language_info":{"name":"python","version":"3.6.4","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"}},"nbformat":4,"nbformat_minor":1}