{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.11.11","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"nvidiaTeslaT4","dataSources":[{"sourceId":104884,"sourceType":"datasetVersion","datasetId":54339}],"dockerImageVersionId":31040,"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":true}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","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\n\nimport numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n\n# Input data files are available in the read-only \"../input/\" directory\n# For example, running this (by clicking run or pressing Shift+Enter) will list all files under the input directory\n\nimport os\nfor dirname, _, filenames in os.walk('/kaggle/input'):\n for filename in filenames:\n print(os.path.join(dirname, filename))\n\n# You can write up to 20GB to the current directory (/kaggle/working/) that gets preserved as output when you create a version using \"Save & Run All\" \n# You can also write temporary files to /kaggle/temp/, but they won't be saved outside of the current session","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"# Importing Libraries","metadata":{}},{"cell_type":"code","source":"#1st step - Importing all the required Libraries\nimport pandas as pd\nimport numpy as np\nimport os\nimport matplotlib.pyplot as plt\nimport seaborn as sns\n\nimport torch\nimport torchvision\nfrom torchvision import datasets, transforms, models\nfrom torch.utils.data import DataLoader, Dataset\nimport torch.nn as nn\nimport torch.optim as optim\n\nfrom sklearn.metrics import accuracy_score, confusion_matrix, classification_report\nfrom PIL import Image","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-06-19T16:40:32.917057Z","iopub.execute_input":"2025-06-19T16:40:32.917369Z","iopub.status.idle":"2025-06-19T16:40:32.921831Z","shell.execute_reply.started":"2025-06-19T16:40:32.917348Z","shell.execute_reply":"2025-06-19T16:40:32.921035Z"}},"outputs":[],"execution_count":10},{"cell_type":"markdown","source":"# Loading the Dataset","metadata":{}},{"cell_type":"code","source":"#Loading the dataset\ndata = pd.read_csv(\"/kaggle/input/skin-cancer-mnist-ham10000/HAM10000_metadata.csv\")\ndata.head()","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-06-19T16:40:38.950796Z","iopub.execute_input":"2025-06-19T16:40:38.95106Z","iopub.status.idle":"2025-06-19T16:40:38.976913Z","shell.execute_reply.started":"2025-06-19T16:40:38.951043Z","shell.execute_reply":"2025-06-19T16:40:38.976348Z"}},"outputs":[{"execution_count":11,"output_type":"execute_result","data":{"text/plain":" lesion_id image_id dx dx_type age sex localization\n0 HAM_0000118 ISIC_0027419 bkl histo 80.0 male scalp\n1 HAM_0000118 ISIC_0025030 bkl histo 80.0 male scalp\n2 HAM_0002730 ISIC_0026769 bkl histo 80.0 male scalp\n3 HAM_0002730 ISIC_0025661 bkl histo 80.0 male scalp\n4 HAM_0001466 ISIC_0031633 bkl histo 75.0 male ear","text/html":"
\n\n\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n
lesion_idimage_iddxdx_typeagesexlocalization
0HAM_0000118ISIC_0027419bklhisto80.0malescalp
1HAM_0000118ISIC_0025030bklhisto80.0malescalp
2HAM_0002730ISIC_0026769bklhisto80.0malescalp
3HAM_0002730ISIC_0025661bklhisto80.0malescalp
4HAM_0001466ISIC_0031633bklhisto75.0maleear
\n
"},"metadata":{}}],"execution_count":11},{"cell_type":"markdown","source":"# Filtering the dataset to only include 3 labels ","metadata":{}},{"cell_type":"code","source":"#filtering the dataset to only include mel,nv,bkl labels\nchosen_classes = ['mel','nv', 'bkl']\ndata = data[data['dx'].isin(chosen_classes)]","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-06-19T16:41:53.723271Z","iopub.execute_input":"2025-06-19T16:41:53.723815Z","iopub.status.idle":"2025-06-19T16:41:53.729775Z","shell.execute_reply.started":"2025-06-19T16:41:53.723789Z","shell.execute_reply":"2025-06-19T16:41:53.729125Z"}},"outputs":[],"execution_count":12},{"cell_type":"code","source":"#map the labels to numeric values\nlabel_map = {'mel':0, 'nv':1, 'bkl':2}\ndata['label'] = data['dx'].map(label_map)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-06-19T16:41:57.226793Z","iopub.execute_input":"2025-06-19T16:41:57.227104Z","iopub.status.idle":"2025-06-19T16:41:57.232797Z","shell.execute_reply.started":"2025-06-19T16:41:57.227053Z","shell.execute_reply":"2025-06-19T16:41:57.232138Z"}},"outputs":[],"execution_count":13},{"cell_type":"markdown","source":"# Dividing the Dataset","metadata":{}},{"cell_type":"code","source":"#dividing dataset to training, testing, and validation data\nfrom sklearn.model_selection import train_test_split\ntrain_data, temp_data = train_test_split(data, test_size = 0.3, stratify = data['label'], random_state = 42)\nvalidation_data, test_data = train_test_split(temp_data, test_size = 0.5, stratify = temp_data['label'], random_state = 42 )","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-06-19T16:42:00.292846Z","iopub.execute_input":"2025-06-19T16:42:00.293524Z","iopub.status.idle":"2025-06-19T16:42:00.308934Z","shell.execute_reply.started":"2025-06-19T16:42:00.293491Z","shell.execute_reply":"2025-06-19T16:42:00.308052Z"}},"outputs":[],"execution_count":14},{"cell_type":"markdown","source":"# Data Preprocessing","metadata":{}},{"cell_type":"code","source":"#adding a file path column\ntrain_data['filename'] = train_data['image_id'] + '.jpg'\ntest_data['filename'] = test_data['image_id'] + '.jpg'\nvalidation_data['filename'] = validation_data['image_id'] + '.jpg'","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-06-19T16:42:03.299875Z","iopub.execute_input":"2025-06-19T16:42:03.30019Z","iopub.status.idle":"2025-06-19T16:42:03.307159Z","shell.execute_reply.started":"2025-06-19T16:42:03.300167Z","shell.execute_reply":"2025-06-19T16:42:03.30658Z"}},"outputs":[],"execution_count":15},{"cell_type":"code","source":"from tensorflow.keras.preprocessing.image import ImageDataGenerator\n\ntrain_mod = ImageDataGenerator(rescale = 1./255, rotation_range = 16, horizontal_flip = True, zoom_range = 0.1)\n \n\ntest_mod = ImageDataGenerator(rescale = 1./255)\nval_mod = ImageDataGenerator(rescale = 1./255)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-06-19T16:42:07.067609Z","iopub.execute_input":"2025-06-19T16:42:07.067867Z","iopub.status.idle":"2025-06-19T16:42:07.072231Z","shell.execute_reply.started":"2025-06-19T16:42:07.067848Z","shell.execute_reply":"2025-06-19T16:42:07.071638Z"}},"outputs":[],"execution_count":16},{"cell_type":"code","source":"import os\nimport shutil\nsource1 = '/kaggle/input/skin-cancer-mnist-ham10000/HAM10000_images_part_1'\nsource2 = '/kaggle/input/skin-cancer-mnist-ham10000/HAM10000_images_part_2'\ndestination = 'HAM10000_images'\n\nos.makedirs(destination, exist_ok = True)\nfor folder in [source1, source2]:\n for file in os.listdir(folder):\n if file.endswith('.jpg'):\n shutil.copy(os.path.join(folder,file), os.path.join(destination,file))\n","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-06-19T16:42:10.54063Z","iopub.execute_input":"2025-06-19T16:42:10.541403Z","iopub.status.idle":"2025-06-19T16:44:24.199552Z","shell.execute_reply.started":"2025-06-19T16:42:10.54136Z","shell.execute_reply":"2025-06-19T16:44:24.198921Z"}},"outputs":[],"execution_count":17},{"cell_type":"code","source":"#creating generators\n\ntrain_gen = train_mod.flow_from_dataframe(dataframe = train_data, \ndirectory = 'HAM10000_images/',\nx_col = 'filename',\ny_col = 'dx',\ntarget_size = (224,224),\nbatch_size = 32,\nclass_mode = 'categorical'\n )\n\nval_gen = val_mod.flow_from_dataframe(dataframe = validation_data, \ndirectory = 'HAM10000_images/',\nx_col = 'filename',\ny_col = 'dx',\ntarget_size = (224,224),\nbatch_size = 32,\nclass_mode = 'categorical'\n )\ntest_gen = test_mod.flow_from_dataframe(dataframe = test_data, \ndirectory = 'HAM10000_images/',\nx_col = 'filename',\ny_col = 'dx',\ntarget_size = (224,224),\nbatch_size = 32,\nclass_mode = 'categorical'\n )\n","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-06-19T16:44:31.028326Z","iopub.execute_input":"2025-06-19T16:44:31.029001Z","iopub.status.idle":"2025-06-19T16:44:31.10802Z","shell.execute_reply.started":"2025-06-19T16:44:31.028973Z","shell.execute_reply":"2025-06-19T16:44:31.107217Z"}},"outputs":[{"name":"stdout","text":"Found 6241 validated image filenames belonging to 3 classes.\nFound 1338 validated image filenames belonging to 3 classes.\nFound 1338 validated image filenames belonging to 3 classes.\n","output_type":"stream"}],"execution_count":18},{"cell_type":"code","source":"import tensorflow as tf\nfrom tensorflow.keras.applications import ResNet50\nfrom tensorflow.keras.models import Model\nfrom tensorflow.keras.layers import Dense, GlobalAveragePooling2D, Dropout\nfrom tensorflow.keras.optimizers import Adam\nfrom sklearn.metrics import classification_report, confusion_matrix\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport seaborn as sns","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-06-19T16:44:37.187199Z","iopub.execute_input":"2025-06-19T16:44:37.187777Z","iopub.status.idle":"2025-06-19T16:44:37.202527Z","shell.execute_reply.started":"2025-06-19T16:44:37.187757Z","shell.execute_reply":"2025-06-19T16:44:37.201883Z"}},"outputs":[],"execution_count":19},{"cell_type":"markdown","source":"# Model Training","metadata":{}},{"cell_type":"code","source":"model_base = ResNet50(weights = 'imagenet', include_top = False, input_shape = (224,224,3))\n#Removing top layers\nmodel_base.Trainable = False\n\n#Adding custom Layers\nnew_layer = model_base.output\nnew_layer = GlobalAveragePooling2D()(new_layer)\nnew_layer = Dropout(0.3)(new_layer)\nfinal_predictions = Dense(3, activation = 'softmax')(new_layer)\n\nfinal_model = Model(inputs = model_base.input, outputs = final_predictions )\n\n#compiling the model\nfinal_model.compile(optimizer = Adam(), loss = 'categorical_crossentropy', metrics = ['accuracy'])\n\n#training the model\nhistory = final_model.fit(train_gen, \n validation_data = val_gen,\n epochs = 10)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-06-19T16:44:42.171048Z","iopub.execute_input":"2025-06-19T16:44:42.171596Z","iopub.status.idle":"2025-06-19T17:05:46.531849Z","shell.execute_reply.started":"2025-06-19T16:44:42.171567Z","shell.execute_reply":"2025-06-19T17:05:46.531273Z"}},"outputs":[{"name":"stderr","text":"I0000 00:00:1750351482.757115 35 gpu_device.cc:2022] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 13942 MB memory: -> device: 0, name: Tesla T4, pci bus id: 0000:00:04.0, compute capability: 7.5\nI0000 00:00:1750351482.757998 35 gpu_device.cc:2022] Created device /job:localhost/replica:0/task:0/device:GPU:1 with 13942 MB memory: -> device: 1, name: Tesla T4, pci bus id: 0000:00:05.0, compute capability: 7.5\n","output_type":"stream"},{"name":"stdout","text":"Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/resnet/resnet50_weights_tf_dim_ordering_tf_kernels_notop.h5\n\u001b[1m94765736/94765736\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 0us/step\n","output_type":"stream"},{"name":"stderr","text":"/usr/local/lib/python3.11/dist-packages/keras/src/trainers/data_adapters/py_dataset_adapter.py:121: UserWarning: Your `PyDataset` class should call `super().__init__(**kwargs)` in its constructor. `**kwargs` can include `workers`, `use_multiprocessing`, `max_queue_size`. Do not pass these arguments to `fit()`, as they will be ignored.\n self._warn_if_super_not_called()\n","output_type":"stream"},{"name":"stdout","text":"Epoch 1/10\n","output_type":"stream"},{"name":"stderr","text":"WARNING: All log messages before absl::InitializeLog() is called are written to STDERR\nI0000 00:00:1750351539.571600 100 service.cc:148] XLA service 0x78b5c8001c40 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:\nI0000 00:00:1750351539.573592 100 service.cc:156] StreamExecutor device (0): Tesla T4, Compute Capability 7.5\nI0000 00:00:1750351539.573616 100 service.cc:156] StreamExecutor device (1): Tesla T4, Compute Capability 7.5\nI0000 00:00:1750351544.677549 100 cuda_dnn.cc:529] Loaded cuDNN version 90300\nI0000 00:00:1750351568.665769 100 device_compiler.h:188] Compiled cluster using XLA! This line is logged at most once for the lifetime of the process.\n","output_type":"stream"},{"name":"stdout","text":"\u001b[1m196/196\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m226s\u001b[0m 744ms/step - accuracy: 0.7572 - loss: 0.6662 - val_accuracy: 0.7519 - val_loss: 0.8711\nEpoch 2/10\n\u001b[1m196/196\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m117s\u001b[0m 593ms/step - accuracy: 0.7800 - loss: 0.5348 - val_accuracy: 0.7519 - val_loss: 0.9229\nEpoch 3/10\n\u001b[1m196/196\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m114s\u001b[0m 583ms/step - accuracy: 0.7811 - loss: 0.5548 - val_accuracy: 0.7519 - val_loss: 125.5597\nEpoch 4/10\n\u001b[1m196/196\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m116s\u001b[0m 590ms/step - accuracy: 0.7831 - loss: 0.5565 - val_accuracy: 0.7070 - val_loss: 1.0213\nEpoch 5/10\n\u001b[1m196/196\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m116s\u001b[0m 591ms/step - accuracy: 0.7929 - loss: 0.5083 - val_accuracy: 0.6644 - val_loss: 0.9804\nEpoch 6/10\n\u001b[1m196/196\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m113s\u001b[0m 575ms/step - accuracy: 0.8053 - loss: 0.5078 - val_accuracy: 0.7175 - val_loss: 0.6906\nEpoch 7/10\n\u001b[1m196/196\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m114s\u001b[0m 580ms/step - accuracy: 0.7990 - loss: 0.5098 - val_accuracy: 0.5561 - val_loss: 0.9290\nEpoch 8/10\n\u001b[1m196/196\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m113s\u001b[0m 575ms/step - accuracy: 0.7901 - loss: 0.6143 - val_accuracy: 0.7683 - val_loss: 0.7824\nEpoch 9/10\n\u001b[1m196/196\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m114s\u001b[0m 581ms/step - accuracy: 0.7770 - loss: 0.5399 - val_accuracy: 0.7945 - val_loss: 0.5201\nEpoch 10/10\n\u001b[1m196/196\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m115s\u001b[0m 586ms/step - accuracy: 0.8027 - loss: 0.4935 - val_accuracy: 0.8027 - val_loss: 0.5020\n","output_type":"stream"}],"execution_count":20},{"cell_type":"markdown","source":"# Model Evaluation","metadata":{}},{"cell_type":"code","source":"loss, accuracy = final_model.evaluate(test_gen)\nprint(\"accuracy is\", accuracy)\n\ntrue_output = test_gen.classes\npredicted_output = final_model.predict(test_gen)\nprediction = np.argmax(predicted_output, axis = 1)\n\nlabels = list(test_gen.class_indices.keys())\nprint(\"Classification Report\")\nprint(classification_report(true_output, prediction, target_names = labels))\n\ncm = confusion_matrix(true_output, prediction)\nplt.figure(figsize=(5,4))\n\nsns.heatmap(cm, annot = True, fmt='d', cmap = 'Blues', xticklabels = labels, yticklabels = labels)\nplt.title('Confusion Matrix')\nplt.xlabel('Predicted')\nplt.ylabel('Actual')\nplt.show()","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2025-06-19T17:27:10.694282Z","iopub.execute_input":"2025-06-19T17:27:10.694901Z","iopub.status.idle":"2025-06-19T17:27:36.89391Z","shell.execute_reply.started":"2025-06-19T17:27:10.694878Z","shell.execute_reply":"2025-06-19T17:27:36.893216Z"}},"outputs":[{"name":"stdout","text":"\u001b[1m42/42\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 200ms/step - accuracy: 0.7965 - loss: 0.4790\naccuracy is 0.8064274787902832\n\u001b[1m42/42\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 293ms/step\nClassification Report\n precision recall f1-score support\n\n bkl 0.15 0.18 0.16 165\n mel 0.17 0.04 0.07 167\n nv 0.76 0.83 0.79 1006\n\n accuracy 0.65 1338\n macro avg 0.36 0.35 0.34 1338\nweighted avg 0.61 0.65 0.63 1338\n\n","output_type":"stream"},{"output_type":"display_data","data":{"text/plain":"
","image/png":"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\n"},"metadata":{}}],"execution_count":21},{"cell_type":"markdown","source":"# Model Summary and Evaluation ","metadata":{}},{"cell_type":"markdown","source":"**Model Summary**- In this project, I used the pre-trained model ResNet50, as it is a very powerful model for image classification tasks. First, I removed the top layer, and then added new layers - Pooling layer, dropout layer so as to prevent overfitting, and finally the dense layer with softmax as activation function. Softmax is used for multiclass classification. For training the model - categorical cross entropy is used as loss function, and adam optimizer is used. Preprocessing is done by ImageDataGenerator. \n\n**Model Evaluation** - After training the model, we tested the model using the test set. The accuracy is 80.64% and loss is 0.48. The confusion matrix also displays that model is able to predict well. Nevus class showed the highest accuracy, and the model performed fairly for rest of the classes. Hence we can say that transfer learning is a good technique for skin condition classification, and can perform better with fine-tuning and by using more balanced dataset. ","metadata":{}}]}