2.1 | Load Dataset
\ntrain_labels.csv, it contains a reference to an image ID (id) & its classification allocation (label)","metadata":{}},{"cell_type":"code","source":"labels_df = pd.read_csv('/kaggle/input/histopathologic-cancer-detection/train_labels.csv')\nprint(labels_df.head().to_markdown())","metadata":{"execution":{"iopub.status.busy":"2023-07-02T19:33:13.475442Z","iopub.execute_input":"2023-07-02T19:33:13.475759Z","iopub.status.idle":"2023-07-02T19:33:13.978762Z","shell.execute_reply.started":"2023-07-02T19:33:13.475732Z","shell.execute_reply":"2023-07-02T19:33:13.977753Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"os.listdir('/kaggle/input/histopathologic-cancer-detection/')","metadata":{"execution":{"iopub.status.busy":"2023-07-02T19:33:13.980257Z","iopub.execute_input":"2023-07-02T19:33:13.980575Z","iopub.status.idle":"2023-07-02T19:33:13.987093Z","shell.execute_reply.started":"2023-07-02T19:33:13.980548Z","shell.execute_reply":"2023-07-02T19:33:13.986160Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"labels_df.shape","metadata":{"execution":{"iopub.status.busy":"2023-07-02T19:33:13.988245Z","iopub.execute_input":"2023-07-02T19:33:13.988523Z","iopub.status.idle":"2023-07-02T19:33:13.998656Z","shell.execute_reply.started":"2023-07-02T19:33:13.988498Z","shell.execute_reply":"2023-07-02T19:33:13.997761Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"2.2 | Check for duplicate entries
\n2.3 | Target feature class balance
\n2.4 | Dataset preview
\n3.1 | Custom dataset class
\nDataset class by subclassing the Pytorch Dataset class:\n - We need just two essential fuctions __len__ & __getitem__ in our custom class \n- To speed up the training process, were using only 4000 samples for the entire dataset (","metadata":{}},{"cell_type":"code","source":"torch.manual_seed(0) # fix random seed\n\nclass pytorch_data(Dataset):\n \n def __init__(self,data_dir,transform,data_type=\"train\"): \n \n # Get Image File Names\n cdm_data=os.path.join(data_dir,data_type) # directory of files\n \n file_names = os.listdir(cdm_data) # get list of images in that directory \n idx_choose = np.random.choice(np.arange(len(file_names)), \n 4000,\n replace=False).tolist()\n file_names_sample = [file_names[x] for x in idx_choose]\n self.full_filenames = [os.path.join(cdm_data, f) for f in file_names_sample] # get the full path to images\n \n # Get Labels\n labels_data=os.path.join(data_dir,\"train_labels.csv\") \n labels_df=pd.read_csv(labels_data)\n labels_df.set_index(\"id\", inplace=True) # set data frame index to id\n self.labels = [labels_df.loc[filename[:-4]].values[0] for filename in file_names_sample] # obtained labels from df\n self.transform = transform\n \n def __len__(self):\n return len(self.full_filenames) # size of dataset\n \n def __getitem__(self, idx):\n # open image, apply transforms and return with label\n image = Image.open(self.full_filenames[idx]) # Open Image with PIL\n image = self.transform(image) # Apply Specific Transformation to Image\n return image, self.labels[idx]\n","metadata":{"execution":{"iopub.status.busy":"2023-07-02T19:33:21.558710Z","iopub.execute_input":"2023-07-02T19:33:21.559036Z","iopub.status.idle":"2023-07-02T19:33:21.573767Z","shell.execute_reply.started":"2023-07-02T19:33:21.559008Z","shell.execute_reply":"2023-07-02T19:33:21.572873Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# define transformation that converts a PIL image into PyTorch tensors\nimport torchvision.transforms as transforms\ndata_transformer = transforms.Compose([transforms.ToTensor(),\n transforms.Resize((46,46))])","metadata":{"execution":{"iopub.status.busy":"2023-07-02T19:33:21.574919Z","iopub.execute_input":"2023-07-02T19:33:21.575206Z","iopub.status.idle":"2023-07-02T19:33:21.587773Z","shell.execute_reply.started":"2023-07-02T19:33:21.575182Z","shell.execute_reply":"2023-07-02T19:33:21.586802Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Define an object of the custom dataset for the train folder.\ndata_dir = '/kaggle/input/histopathologic-cancer-detection/'\nimg_dataset = pytorch_data(data_dir, data_transformer, \"train\") # Histopathalogic images","metadata":{"execution":{"iopub.status.busy":"2023-07-02T19:33:21.589145Z","iopub.execute_input":"2023-07-02T19:33:21.589493Z","iopub.status.idle":"2023-07-02T19:33:25.639712Z","shell.execute_reply.started":"2023-07-02T19:33:21.589467Z","shell.execute_reply":"2023-07-02T19:33:25.638620Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# load an example tensor\nimg,label=img_dataset[10]\nprint(img.shape,torch.min(img),torch.max(img))","metadata":{"execution":{"iopub.status.busy":"2023-07-02T19:33:25.641063Z","iopub.execute_input":"2023-07-02T19:33:25.641390Z","iopub.status.idle":"2023-07-02T19:33:25.665235Z","shell.execute_reply.started":"2023-07-02T19:33:25.641362Z","shell.execute_reply":"2023-07-02T19:33:25.664340Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## 4 | Splitting the Dataset \n\n4.1 | random_split
\n4.2 | Training subset examples
\n4.3 | Validation subset examples
\nnn.Module\n","metadata":{}},{"cell_type":"code","source":"def findConv2dOutShape(hin,win,conv,pool=2):\n # get conv arguments\n kernel_size=conv.kernel_size\n stride=conv.stride\n padding=conv.padding\n dilation=conv.dilation\n\n hout=np.floor((hin+2*padding[0]-dilation[0]*(kernel_size[0]-1)-1)/stride[0]+1)\n wout=np.floor((win+2*padding[1]-dilation[1]*(kernel_size[1]-1)-1)/stride[1]+1)\n\n if pool:\n hout/=pool\n wout/=pool\n return int(hout),int(wout)\n\nimport torch.nn as nn\nimport torch.nn.functional as F\n\n# Neural Network\nclass Network(nn.Module):\n \n # Network Initialisation\n def __init__(self, params):\n \n super(Network, self).__init__()\n \n Cin,Hin,Win=params[\"shape_in\"]\n init_f=params[\"initial_filters\"] \n num_fc1=params[\"num_fc1\"] \n num_classes=params[\"num_classes\"] \n self.dropout_rate=params[\"dropout_rate\"] \n \n # Convolution Layers\n self.conv1 = nn.Conv2d(Cin, init_f, kernel_size=3)\n h,w=findConv2dOutShape(Hin,Win,self.conv1)\n self.conv2 = nn.Conv2d(init_f, 2*init_f, kernel_size=3)\n h,w=findConv2dOutShape(h,w,self.conv2)\n self.conv3 = nn.Conv2d(2*init_f, 4*init_f, kernel_size=3)\n h,w=findConv2dOutShape(h,w,self.conv3)\n self.conv4 = nn.Conv2d(4*init_f, 8*init_f, kernel_size=3)\n h,w=findConv2dOutShape(h,w,self.conv4)\n \n # compute the flatten size\n self.num_flatten=h*w*8*init_f\n self.fc1 = nn.Linear(self.num_flatten, num_fc1)\n self.fc2 = nn.Linear(num_fc1, num_classes)\n\n def forward(self,X):\n \n # Convolution & Pool Layers\n X = F.relu(self.conv1(X)); \n X = F.max_pool2d(X, 2, 2)\n X = F.relu(self.conv2(X))\n X = F.max_pool2d(X, 2, 2)\n X = F.relu(self.conv3(X))\n X = F.max_pool2d(X, 2, 2)\n X = F.relu(self.conv4(X))\n X = F.max_pool2d(X, 2, 2)\n\n X = X.view(-1, self.num_flatten)\n \n X = F.relu(self.fc1(X))\n X=F.dropout(X, self.dropout_rate)\n X = self.fc2(X)\n return F.log_softmax(X, dim=1)","metadata":{"_kg_hide-input":false,"execution":{"iopub.status.busy":"2023-07-02T19:34:32.579952Z","iopub.execute_input":"2023-07-02T19:34:32.580959Z","iopub.status.idle":"2023-07-02T19:34:32.600212Z","shell.execute_reply.started":"2023-07-02T19:34:32.580916Z","shell.execute_reply":"2023-07-02T19:34:32.599183Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Neural Network Predefined Parameters\nparams_model={\n \"shape_in\": (3,46,46), \n \"initial_filters\": 8, \n \"num_fc1\": 100,\n \"dropout_rate\": 0.25,\n \"num_classes\": 2}\n\n# Create instantiation of Network class\ncnn_model = Network(params_model)\n\n# define computation hardware approach (GPU/CPU)\ndevice = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\nmodel = cnn_model.to(device)","metadata":{"execution":{"iopub.status.busy":"2023-07-02T19:34:32.845167Z","iopub.execute_input":"2023-07-02T19:34:32.845491Z","iopub.status.idle":"2023-07-02T19:34:32.854694Z","shell.execute_reply.started":"2023-07-02T19:34:32.845464Z","shell.execute_reply":"2023-07-02T19:34:32.853661Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"from torchsummary import summary\nsummary(cnn_model, input_size=(3, 46, 46),device=device.type)","metadata":{"execution":{"iopub.status.busy":"2023-07-02T19:34:34.112307Z","iopub.execute_input":"2023-07-02T19:34:34.112759Z","iopub.status.idle":"2023-07-02T19:34:34.136433Z","shell.execute_reply.started":"2023-07-02T19:34:34.112729Z","shell.execute_reply":"2023-07-02T19:34:34.135072Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## 8 | Loss Function Definition \n\n- Loss Functions are one of the key pieces of an effective deep learning solution.\n- Pytorch uses loss functions to determine how it will update the network to reach the desired solution.\n- The standard loss function for classification tasks is __cross entropy loss__ or __logloss__\n- When defining a loss function, we need to consider, the number of model outputs and their activation functions.\n- For binary classification tasks, we can choose one or two outputs.\n- It is recommended to use __log_softmax__ as it is easier to expand to multiclass classification; PyTorch combines the log and softmax operations into one function, due to numerical stability and speed.","metadata":{}},{"cell_type":"code","source":"loss_func = nn.NLLLoss(reduction=\"sum\")","metadata":{"execution":{"iopub.status.busy":"2023-07-02T19:34:37.615984Z","iopub.execute_input":"2023-07-02T19:34:37.616438Z","iopub.status.idle":"2023-07-02T19:34:37.621585Z","shell.execute_reply.started":"2023-07-02T19:34:37.616404Z","shell.execute_reply":"2023-07-02T19:34:37.620385Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## 9 | Optimiser Definition \n\n- Training the network involves passing data through the network:\n - Using the **loss function** to **determine the difference between the prediction & true value**\n - Which is then followed by using of that info to **update the weights** of the network \n - In an attempt to **make the loss function return as small of a loss as possible, performing updates on the neural network**, an **optimiser** is used\n- The torch.optim contains implementations of common optimisers\n- The **optimiser** will **hold the current state and will update the parameters based on the computed gradients**\n- For binary classification taskss, __SGD__, __Adam__ Optimisers are commonly used, let's use the latter here.","metadata":{}},{"cell_type":"code","source":"from torch import optim\nopt = optim.Adam(cnn_model.parameters(), lr=3e-4)\nlr_scheduler = ReduceLROnPlateau(opt, mode='min',factor=0.5, patience=20,verbose=1)","metadata":{"execution":{"iopub.status.busy":"2023-07-02T19:34:39.627355Z","iopub.execute_input":"2023-07-02T19:34:39.627791Z","iopub.status.idle":"2023-07-02T19:34:39.634050Z","shell.execute_reply.started":"2023-07-02T19:34:39.627759Z","shell.execute_reply":"2023-07-02T19:34:39.633024Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## 10 | Training Model \n\n10.1 | Helper functions
\ntrain_val will utiliser three functions:\n- get_lr : get the learning rate as it is adjusted \n- loss_batch : get the loss value for the particular batch\n- loss_epoch : get the entire loss for an epoch iteration","metadata":{}},{"cell_type":"code","source":"''' Helper Functions'''\n\n# Function to get the learning rate\ndef get_lr(opt):\n for param_group in opt.param_groups:\n return param_group['lr']\n\n# Function to compute the loss value per batch of data\ndef loss_batch(loss_func, output, target, opt=None):\n \n loss = loss_func(output, target) # get loss\n pred = output.argmax(dim=1, keepdim=True) # Get Output Class\n metric_b=pred.eq(target.view_as(pred)).sum().item() # get performance metric\n \n if opt is not None:\n opt.zero_grad()\n loss.backward()\n opt.step()\n\n return loss.item(), metric_b\n\n# Compute the loss value & performance metric for the entire dataset (epoch)\ndef loss_epoch(model,loss_func,dataset_dl,opt=None):\n \n run_loss=0.0 \n t_metric=0.0\n len_data=len(dataset_dl.dataset)\n\n # internal loop over dataset\n for xb, yb in dataset_dl:\n # move batch to device\n xb=xb.to(device)\n yb=yb.to(device)\n output=model(xb) # get model output\n loss_b,metric_b=loss_batch(loss_func, output, yb, opt) # get loss per batch\n run_loss+=loss_b # update running loss\n\n if metric_b is not None: # update running metric\n t_metric+=metric_b \n \n loss=run_loss/float(len_data) # average loss value\n metric=t_metric/float(len_data) # average metric value\n \n return loss, metric","metadata":{"_kg_hide-input":true,"execution":{"iopub.status.busy":"2023-07-02T19:34:44.602224Z","iopub.execute_input":"2023-07-02T19:34:44.603481Z","iopub.status.idle":"2023-07-02T19:34:44.614585Z","shell.execute_reply.started":"2023-07-02T19:34:44.603440Z","shell.execute_reply":"2023-07-02T19:34:44.613513Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"params_train={\n \"train\": train_dl,\"val\": val_dl,\n \"epochs\": 50,\n \"optimiser\": optim.Adam(cnn_model.parameters(),\n lr=3e-4),\n \"lr_change\": ReduceLROnPlateau(opt,\n mode='min',\n factor=0.5,\n patience=20,\n verbose=0),\n \"f_loss\": nn.NLLLoss(reduction=\"sum\"),\n \"weight_path\": \"weights.pt\",\n \"check\": False, \n}\n","metadata":{"execution":{"iopub.status.busy":"2023-07-02T19:34:46.787371Z","iopub.execute_input":"2023-07-02T19:34:46.787801Z","iopub.status.idle":"2023-07-02T19:34:46.794696Z","shell.execute_reply.started":"2023-07-02T19:34:46.787769Z","shell.execute_reply":"2023-07-02T19:34:46.793767Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"10.2 | Main training function
\n10.3 | Training Process
\n