{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"## Importing Libraries","metadata":{"id":"KZLvfg_GMCBz"}},{"cell_type":"code","source":"import os\n\nimport pandas as pd\nimport numpy as np\nimport tensorflow as tf\nimport pickle\nimport matplotlib.pyplot as plt\nimport seaborn as sns\nimport random as rnd\nimport cv2\n\nfrom keras import layers\nfrom keras.models import Sequential\nfrom keras.preprocessing import image\nfrom keras.applications.imagenet_utils import preprocess_input\nfrom keras.layers import Input, Dense, Activation, Dropout\nfrom keras.layers import Flatten, BatchNormalization, Conv2D\nfrom keras.layers import MaxPooling2D, AveragePooling2D\n\n\nfrom PIL import Image\nfrom tqdm import tqdm\n\n!pip install livelossplot\nfrom livelossplot import PlotLossesKeras\n\n%matplotlib inline","metadata":{"_kg_hide-output":true,"id":"dYimI_AlMCB2","execution":{"iopub.status.busy":"2022-04-30T04:32:06.988831Z","iopub.execute_input":"2022-04-30T04:32:06.989375Z","iopub.status.idle":"2022-04-30T04:32:23.712272Z","shell.execute_reply.started":"2022-04-30T04:32:06.989260Z","shell.execute_reply":"2022-04-30T04:32:23.711238Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"tr_df = pd.read_csv('../input/happy-whale-and-dolphin/train.csv')\ntr_df['path'] = '../input/happy-whale-and-dolphin/train_images/' + tr_df['image']\n\npd_df = pd.read_csv('../input/happy-whale-and-dolphin/sample_submission.csv')\npd_df['path'] = '../input/happy-whale-and-dolphin/test_images/' + pd_df['image']","metadata":{"id":"yis8htrzMCCM","execution":{"iopub.status.busy":"2022-04-30T04:32:23.714829Z","iopub.execute_input":"2022-04-30T04:32:23.715077Z","iopub.status.idle":"2022-04-30T04:32:23.890722Z","shell.execute_reply.started":"2022-04-30T04:32:23.715047Z","shell.execute_reply":"2022-04-30T04:32:23.889798Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Exploratory data analysis","metadata":{"id":"KSZIxHe3NcG0"}},{"cell_type":"code","source":"print(\"Glance at a dataset\")\nprint(tr_df.head())\n\nprint(\"columns of the dataset are as follows\")\nprint(tr_df.columns)\n\nprint(\"Samples count of the training data:\")\nprint(len(tr_df))\n\nprint(\"Species count in the training data\")\ntr_df.species.value_counts()","metadata":{"id":"npJdPQAaMCCX","execution":{"iopub.status.busy":"2022-04-30T04:32:23.891877Z","iopub.execute_input":"2022-04-30T04:32:23.892121Z","iopub.status.idle":"2022-04-30T04:32:23.913606Z","shell.execute_reply.started":"2022-04-30T04:32:23.892094Z","shell.execute_reply":"2022-04-30T04:32:23.913033Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Data Cleaning\n### After looking at the species we can see some duplicate values presenting similar entity\n* `bottlenose_dolpin` and `bottlenose_dolphin`\n* `kiler_whale` and `killer_whale`\n* `beluga` and `beluga_whale`\n\n### Due to extreme similarities between this two species we combined them \n* `globis` and `pilot_whale` into => `short_finned_pilot_whale`","metadata":{"id":"5vwgYxxWMCCd"}},{"cell_type":"code","source":"print(\"Number of unique species, Before fixing duplicate labels: \", tr_df['species'].nunique())\n\ntr_df['species'].replace({'bottlenose_dolpin' : 'bottlenose_dolphin', 'kiler_whale' : 'killer_whale', 'beluga' : 'beluga_whale', 'globis' : 'short_finned_pilot_whale', 'pilot_whale' : 'short_finned_pilot_whale'}, inplace =True)\n\nprint('After data cleaning steps')\nprint(\"Total number of unique species: \", tr_df['species'].nunique())\n\ntr_df['class'] = tr_df['species'].apply(lambda x: x.split('_')[-1])\n","metadata":{"id":"OHkcLAZFMCCg","execution":{"iopub.status.busy":"2022-04-30T04:32:23.914980Z","iopub.execute_input":"2022-04-30T04:32:23.915756Z","iopub.status.idle":"2022-04-30T04:32:23.960254Z","shell.execute_reply.started":"2022-04-30T04:32:23.915722Z","shell.execute_reply":"2022-04-30T04:32:23.958886Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# glance at data after performing these steps.\ntr_df.head()","metadata":{"id":"ZE_q0bppY-y4","execution":{"iopub.status.busy":"2022-04-30T04:32:23.961848Z","iopub.execute_input":"2022-04-30T04:32:23.962085Z","iopub.status.idle":"2022-04-30T04:32:23.979304Z","shell.execute_reply.started":"2022-04-30T04:32:23.962060Z","shell.execute_reply":"2022-04-30T04:32:23.978597Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Checking missing data and images","metadata":{"id":"lvLdINenMCCi"}},{"cell_type":"code","source":"# print(tr_df.isna())\n\nprint(tr_df.isna().sum())\nprint(len(os.listdir('../input/happy-whale-and-dolphin/train_images')))","metadata":{"id":"HdyHhOlfMCCk","execution":{"iopub.status.busy":"2022-04-30T04:32:23.981473Z","iopub.execute_input":"2022-04-30T04:32:23.981906Z","iopub.status.idle":"2022-04-30T04:32:24.832793Z","shell.execute_reply.started":"2022-04-30T04:32:23.981856Z","shell.execute_reply":"2022-04-30T04:32:24.831891Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Exploratory Data Analysis \n###by analyzing the distribution of class values","metadata":{"id":"pbasVtBeMCDA"}},{"cell_type":"code","source":"plot = sns.countplot(x = tr_df['class'], color = '#087525')\nsns.despine()\nplot.set_title('Distribution of Class values\\n', font = 'serif', x = 0.1, y=1, fontsize = 14);\nplot.set_ylabel(\"Count\", x = 0.02, font = 'serif', fontsize = 11)\nplot.set_xlabel(\"Species\", fontsize = 11, font = 'serif')\n\nfor p in plot.patches:\n plot.annotate(format(p.get_height(), '.0f'), (p.get_x() + p.get_width() / 2, p.get_height()), \n ha = 'center', va = 'center', xytext = (0, -20),font = 'serif', textcoords = 'offset points', size = 14)","metadata":{"id":"8eoYImoXMCDB","execution":{"iopub.status.busy":"2022-04-30T04:32:24.834359Z","iopub.execute_input":"2022-04-30T04:32:24.834877Z","iopub.status.idle":"2022-04-30T04:32:25.108761Z","shell.execute_reply.started":"2022-04-30T04:32:24.834832Z","shell.execute_reply":"2022-04-30T04:32:25.107588Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"#### Pie chart of percentage of whale and dolphin in the dataset","metadata":{"id":"BAN6peWQMCDB"}},{"cell_type":"code","source":"plt.figure(figsize=(5,5))\ncl_count = tr_df.groupby(['class']).size().reset_index(name = 'counts')\ncolors = sns.color_palette('Paired')[0:9]\nplt.pie(cl_count['counts'], labels=cl_count['class'], colors=colors, autopct='%1.1f%%')\nplt.legend(loc='best')\nplt.show()","metadata":{"id":"2kdRoa0iMCDC","execution":{"iopub.status.busy":"2022-04-30T04:32:25.110267Z","iopub.execute_input":"2022-04-30T04:32:25.110765Z","iopub.status.idle":"2022-04-30T04:32:25.286593Z","shell.execute_reply.started":"2022-04-30T04:32:25.110718Z","shell.execute_reply":"2022-04-30T04:32:25.285413Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"#### Class distribution by each species","metadata":{"id":"EiFZ2_PVMCDD"}},{"cell_type":"code","source":"plt.figure(figsize=(8,8))\nsns.countplot(data=tr_df, y = 'species', palette='crest', dodge=False)\nplt.show()","metadata":{"id":"AwPkIJ36MCDD","execution":{"iopub.status.busy":"2022-04-30T04:32:25.288572Z","iopub.execute_input":"2022-04-30T04:32:25.289292Z","iopub.status.idle":"2022-04-30T04:32:25.670207Z","shell.execute_reply.started":"2022-04-30T04:32:25.289242Z","shell.execute_reply":"2022-04-30T04:32:25.669418Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"#### class distribution by species for both Whale and Dolphin","metadata":{"id":"q5nJA0AEMCDE"}},{"cell_type":"code","source":"fig,ax = plt.subplots(1,2,figsize=(10,5))\n\nwhales = tr_df[tr_df['class']=='whale']\ndolphins = tr_df[tr_df['class']!='whale']\n\nsns.countplot(y=\"species\", data=whales, order=whales.iloc[0:][\"species\"].value_counts().index, ax=ax[0], color = \"#e38f10\")\nax[0].set_title('Whales species Frequency')\nax[0].set_ylabel(None)\n \nsns.countplot(y=\"species\", data=dolphins,order=dolphins.iloc[0:][\"species\"].value_counts().index, ax=ax[1], color = \"#2551a8\")\nax[1].set_title('Dolphins species Frequency')\nax[1].set_ylabel(None)\n\nplt.tight_layout()\nplt.show()","metadata":{"id":"8c6NioBBMCDE","execution":{"iopub.status.busy":"2022-04-30T04:32:25.672673Z","iopub.execute_input":"2022-04-30T04:32:25.672902Z","iopub.status.idle":"2022-04-30T04:32:26.211974Z","shell.execute_reply.started":"2022-04-30T04:32:25.672872Z","shell.execute_reply":"2022-04-30T04:32:26.211164Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"#### Plot the value count graph of each individual","metadata":{"id":"4FdwPQ0eMCDH"}},{"cell_type":"code","source":"tr_df['individual_id'].value_counts().plot()\nplt.xticks(rotation=90)\nplt.show()","metadata":{"id":"OWYyfO7PMCDI","execution":{"iopub.status.busy":"2022-04-30T04:32:26.213176Z","iopub.execute_input":"2022-04-30T04:32:26.214105Z","iopub.status.idle":"2022-04-30T04:32:26.449294Z","shell.execute_reply.started":"2022-04-30T04:32:26.214070Z","shell.execute_reply":"2022-04-30T04:32:26.448452Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"#### Density estimation of each individuals","metadata":{"id":"vwL87pGWMCDI"}},{"cell_type":"code","source":"np.log(tr_df['individual_id'].value_counts()).plot.kde()","metadata":{"id":"LyH1OhIsMCDJ","execution":{"iopub.status.busy":"2022-04-30T04:32:26.450501Z","iopub.execute_input":"2022-04-30T04:32:26.450741Z","iopub.status.idle":"2022-04-30T04:32:27.008176Z","shell.execute_reply.started":"2022-04-30T04:32:26.450711Z","shell.execute_reply":"2022-04-30T04:32:27.007205Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"#### Density estimation of individual for both whale and dolphin","metadata":{"id":"8yvu1N_tMCDK"}},{"cell_type":"code","source":"plt.figure(figsize = (15, 8))\nsns.kdeplot(np.log(tr_df.loc[tr_df['class'] == 'dolphin']['individual_id'].value_counts()))\nsns.kdeplot(np.log(tr_df.loc[tr_df['class'] == 'whale']['individual_id'].value_counts()))\nplt.legend(labels = ['dolphin', 'whale'])\nplt.show()","metadata":{"id":"CNifTN5OMCDK","execution":{"iopub.status.busy":"2022-04-30T04:32:27.009334Z","iopub.execute_input":"2022-04-30T04:32:27.009574Z","iopub.status.idle":"2022-04-30T04:32:27.395427Z","shell.execute_reply.started":"2022-04-30T04:32:27.009544Z","shell.execute_reply":"2022-04-30T04:32:27.394556Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### Counting images based on individual_id","metadata":{"id":"XDvLtDeUMCDN"}},{"cell_type":"code","source":"tr_df['count'] = tr_df.groupby('individual_id',as_index=False)['individual_id'].transform(lambda x: x.count())","metadata":{"id":"AuCi05jNMCDO","execution":{"iopub.status.busy":"2022-04-30T04:32:27.396658Z","iopub.execute_input":"2022-04-30T04:32:27.396894Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"tr_df.head()","metadata":{"id":"iNPJOA4Igq5p","trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"#### Percentage of Individuals with greater than or equal to 5 and less then 21 images","metadata":{"id":"Orry-dtbMCDR"}},{"cell_type":"code","source":"count = 0\nfor i in tr_df['count']:\n if(i > 4 and i <= 20):\n count += 1\nprint(count/len(tr_df))","metadata":{"id":"CMWaFAlyMCDR","trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"\n# Image Resolutions","metadata":{"id":"95pcgo-7MCDT"}},{"cell_type":"code","source":"widths, heights = [], []\n\nfor path in tqdm(tr_df[\"path\"]):\n width, height = Image.open(path).size\n widths.append(width)\n heights.append(height)\n \ntr_df[\"height\"] = heights\ntr_df[\"width\"] = widths\ntr_df[\"dimension\"] = tr_df[\"height\"] * tr_df[\"width\"]","metadata":{"id":"esd_WxNFMCDT","trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### Color Analysis ","metadata":{"id":"pswa7nV3MCDV"}},{"cell_type":"code","source":"def check_greyscale(Img_temp):\n w,h = Img_temp.size\n for i in range(w):\n for j in range(h):\n r,g,b = Img_temp.getpixel((i,j))\n if r != g != b: return False\n return True\n\nsampFr = 0.1\n#get our sampled images\ngrey_list = []\nfor img_name in tr_df['path'].sample(frac=sampFr):\n val = Image.open(img_name).convert('RGB')\n grey_list.append(check_greyscale(val))\nprint(np.sum(grey_list) / len(grey_list))\ndel grey_list","metadata":{"id":"KAU-g9ggMCDd","trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### Get mean intensity for each channel RGB ","metadata":{"id":"cC0DD4cLMCDg"}},{"cell_type":"code","source":"def rgb_get_m(row):\n img = cv2.imread(row['path'])\n img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n return np.sum(img[:,:,0]), np.sum(img[:,:,1]), np.sum(img[:,:,2])\n\ntqdm.pandas()\ntr_df['R'], tr_df['G'], tr_df['B'] = zip(*tr_df.progress_apply(lambda row: rgb_get_m(row), axis=1) )","metadata":{"id":"UTTgyyy6MCDh","trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def col_distribution(df, count):\n fig, axr = plt.subplots(count,2,figsize=(15,15))\n for idx, i in enumerate(np.random.choice(df['path'], count)):\n img = cv2.imread(i)\n img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n axr[idx,0].imshow(img)\n axr[idx,0].axis('off')\n axr[idx,1].set_title('R={:.0f}, G={:.0f}, B={:.0f} '.format(np.mean(img[:,:,0]), np.mean(img[:,:,1]), np.mean(img[:,:,2]))) \n x, y = np.histogram(img[:,:,0], bins=255)\n axr[idx,1].bar(y[:-1], x, label='R', alpha=0.8, color='red')\n x, y = np.histogram(img[:,:,1], bins=255)\n axr[idx,1].bar(y[:-1], x, label='G', alpha=0.8, color='green')\n x, y = np.histogram(img[:,:,2], bins=255)\n axr[idx,1].bar(y[:-1], x, label='B', alpha=0.8, color='blue')\n axr[idx,1].legend()\n axr[idx,1].axis('off')","metadata":{"id":"WjFAX2qWMCDi","trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Color distribution of red images\n\ndf = tr_df[((tr_df['B']*1.05) < tr_df['R']) & ((tr_df['G']*1.05) < tr_df['R'])]\ncol_distribution(df, 8)","metadata":{"id":"z4DlBtJqMCDj","trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Color distribution of Green images\n\ndf = tr_df[(tr_df['G'] > 1.05*tr_df['R']) & (tr_df['G'] > 1.05*tr_df['B'])]\ncol_distribution(df, 8)","metadata":{"id":"vAJaAxxrMCDl","trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Color distribution of Blue images\n\ndf = tr_df[(tr_df['B'] > 1.3*tr_df['R']) & (tr_df['B'] > 1.3*tr_df['G'])]\ncol_distribution(df, 8)","metadata":{"id":"GR7P7sQmMCDk","trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### We used following method for testing the accuracy of our model","metadata":{"id":"2YOWYi0jMCDs"}},{"cell_type":"markdown","source":"We are going to decide the accuracy of our results using the Mean Average Precision @ 5 (MAP@5):\n\n\n\nwhere U -> number of images\nP(k) -> precision at cutoff k\nn -> number of predictions per image\nrel(k) -> indicator function = 1 if item at rank k is a relevant label, 0 otherwise.\n","metadata":{"id":"BXZ5iY9sMCCA"}}]}