{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.10.14","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"nvidiaTeslaT4","dataSources":[{"sourceId":29762,"databundleVersionId":2541532,"sourceType":"competition"}],"dockerImageVersionId":30776,"isInternetEnabled":false,"language":"python","sourceType":"notebook","isGpuEnabled":true}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"import torch\nimport torch.nn as nn\nimport torch.optim as optim\nfrom torch.utils.data import Dataset, DataLoader\nfrom torchvision import transforms\nfrom PIL import Image\nimport pandas as pd\nimport os\nfrom tqdm import tqdm","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2024-10-08T06:02:41.265579Z","iopub.execute_input":"2024-10-08T06:02:41.266295Z","iopub.status.idle":"2024-10-08T06:02:41.272247Z","shell.execute_reply.started":"2024-10-08T06:02:41.266252Z","shell.execute_reply":"2024-10-08T06:02:41.271183Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")","metadata":{"execution":{"iopub.status.busy":"2024-10-08T06:02:41.273917Z","iopub.execute_input":"2024-10-08T06:02:41.274207Z","iopub.status.idle":"2024-10-08T06:02:41.282745Z","shell.execute_reply.started":"2024-10-08T06:02:41.274175Z","shell.execute_reply":"2024-10-08T06:02:41.281700Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Set random seed for reproducibility","metadata":{}},{"cell_type":"code","source":"torch.manual_seed(42)\ntorch.cuda.manual_seed_all(42)","metadata":{"execution":{"iopub.status.busy":"2024-10-08T06:02:41.284642Z","iopub.execute_input":"2024-10-08T06:02:41.285029Z","iopub.status.idle":"2024-10-08T06:02:41.293560Z","shell.execute_reply.started":"2024-10-08T06:02:41.284985Z","shell.execute_reply":"2024-10-08T06:02:41.292710Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Load the training data","metadata":{}},{"cell_type":"code","source":"train_df = pd.read_csv('/kaggle/input/landmark-recognition-2021/train.csv')\ntrain_df['filepath'] = train_df['id'].apply(lambda x: f\"/kaggle/input/landmark-recognition-2021/train/{x[0]}/{x[1]}/{x[2]}/{x}.jpg\")","metadata":{"execution":{"iopub.status.busy":"2024-10-08T06:02:41.294461Z","iopub.execute_input":"2024-10-08T06:02:41.294777Z","iopub.status.idle":"2024-10-08T06:02:43.568790Z","shell.execute_reply.started":"2024-10-08T06:02:41.294722Z","shell.execute_reply":"2024-10-08T06:02:43.567968Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_df = train_df.groupby('landmark_id').apply(lambda x: x.sample(min(len(x), 2))).reset_index(drop=True)","metadata":{"execution":{"iopub.status.busy":"2024-10-08T06:02:43.570827Z","iopub.execute_input":"2024-10-08T06:02:43.571139Z","iopub.status.idle":"2024-10-08T06:03:08.679960Z","shell.execute_reply.started":"2024-10-08T06:02:43.571105Z","shell.execute_reply":"2024-10-08T06:03:08.678837Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"label mapping","metadata":{}},{"cell_type":"code","source":"label_mapping = {label: idx for idx, label in enumerate(train_df['landmark_id'].unique())}\ninv_label_mapping = {v: k for k, v in label_mapping.items()}\ntrain_df['label'] = train_df['landmark_id'].map(label_mapping)\nnum_classes = len(label_mapping)","metadata":{"execution":{"iopub.status.busy":"2024-10-08T06:03:08.681395Z","iopub.execute_input":"2024-10-08T06:03:08.681800Z","iopub.status.idle":"2024-10-08T06:03:09.052459Z","shell.execute_reply.started":"2024-10-08T06:03:08.681760Z","shell.execute_reply":"2024-10-08T06:03:09.051392Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Define the model","metadata":{}},{"cell_type":"code","source":"class LandmarkDataset(Dataset):\n    def __init__(self, image_paths, labels=None, transform=None):\n        self.image_paths = image_paths\n        self.labels = labels\n        self.transform = transform\n\n    def __len__(self):\n        return len(self.image_paths)\n\n    def __getitem__(self, idx):\n        img_path = self.image_paths[idx]\n        image = Image.open(img_path).convert('RGB')\n        \n        if self.transform:\n            image = self.transform(image)\n        \n        if self.labels is not None:\n            return image, self.labels[idx]\n        else:\n            return image","metadata":{"execution":{"iopub.status.busy":"2024-10-08T06:03:09.053975Z","iopub.execute_input":"2024-10-08T06:03:09.054783Z","iopub.status.idle":"2024-10-08T06:03:09.061990Z","shell.execute_reply.started":"2024-10-08T06:03:09.054718Z","shell.execute_reply":"2024-10-08T06:03:09.060971Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"transforms","metadata":{}},{"cell_type":"code","source":"transform = transforms.Compose([\n    transforms.Resize((32, 32)),  \n    transforms.ToTensor(),\n    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])\n])","metadata":{"execution":{"iopub.status.busy":"2024-10-08T06:03:09.063307Z","iopub.execute_input":"2024-10-08T06:03:09.063604Z","iopub.status.idle":"2024-10-08T06:03:09.073797Z","shell.execute_reply.started":"2024-10-08T06:03:09.063570Z","shell.execute_reply":"2024-10-08T06:03:09.072892Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"data loaders","metadata":{}},{"cell_type":"code","source":"train_dataset = LandmarkDataset(train_df['filepath'].values, train_df['label'].values, transform=transform)\ntrain_loader = DataLoader(train_dataset, batch_size=128, shuffle=True, num_workers=4, pin_memory=True)","metadata":{"execution":{"iopub.status.busy":"2024-10-08T06:03:09.074881Z","iopub.execute_input":"2024-10-08T06:03:09.075166Z","iopub.status.idle":"2024-10-08T06:03:09.115798Z","shell.execute_reply.started":"2024-10-08T06:03:09.075135Z","shell.execute_reply":"2024-10-08T06:03:09.114862Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"define model","metadata":{}},{"cell_type":"code","source":"class TinyNet(nn.Module):\n    def __init__(self, num_classes):\n        super(TinyNet, self).__init__()\n        self.features = nn.Sequential(\n            nn.Conv2d(3, 16, 3, 1, 1),\n            nn.ReLU(inplace=True),\n            nn.MaxPool2d(2, 2),\n            nn.Conv2d(16, 32, 3, 1, 1),\n            nn.ReLU(inplace=True),\n            nn.MaxPool2d(2, 2),\n        )\n        self.classifier = nn.Linear(32 * 8 * 8, num_classes)\n\n    def forward(self, x):\n        x = self.features(x)\n        x = torch.flatten(x, 1)\n        x = self.classifier(x)\n        return x","metadata":{"execution":{"iopub.status.busy":"2024-10-08T06:03:09.116901Z","iopub.execute_input":"2024-10-08T06:03:09.117186Z","iopub.status.idle":"2024-10-08T06:03:09.133016Z","shell.execute_reply.started":"2024-10-08T06:03:09.117154Z","shell.execute_reply":"2024-10-08T06:03:09.132084Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model = TinyNet(num_classes).to(device)\ncriterion = nn.CrossEntropyLoss()\noptimizer = optim.Adam(model.parameters(), lr=0.001)","metadata":{"execution":{"iopub.status.busy":"2024-10-08T06:03:09.135668Z","iopub.execute_input":"2024-10-08T06:03:09.135974Z","iopub.status.idle":"2024-10-08T06:03:10.709882Z","shell.execute_reply.started":"2024-10-08T06:03:09.135943Z","shell.execute_reply":"2024-10-08T06:03:10.708810Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Training function","metadata":{}},{"cell_type":"code","source":"def train_epoch(model, loader, criterion, optimizer, device):\n    model.train()\n    running_loss = 0.0\n    correct = 0\n    total = 0\n    for inputs, labels in tqdm(loader, desc=\"Training\"):\n        inputs, labels = inputs.to(device), labels.to(device)\n        optimizer.zero_grad()\n        outputs = model(inputs)\n        loss = criterion(outputs, labels)\n        loss.backward()\n        optimizer.step()\n        running_loss += loss.item() * inputs.size(0)\n        _, predicted = outputs.max(1)\n        total += labels.size(0)\n        correct += predicted.eq(labels).sum().item()\n    return running_loss / len(loader.dataset), correct / total","metadata":{"execution":{"iopub.status.busy":"2024-10-08T06:03:10.711382Z","iopub.execute_input":"2024-10-08T06:03:10.711822Z","iopub.status.idle":"2024-10-08T06:03:10.720512Z","shell.execute_reply.started":"2024-10-08T06:03:10.711773Z","shell.execute_reply":"2024-10-08T06:03:10.719065Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Train the model","metadata":{}},{"cell_type":"code","source":"num_epochs = 4\nfor epoch in range(num_epochs):\n    print(f\"Epoch {epoch+1}/{num_epochs}:\")\n    train_loss, train_acc = train_epoch(model, train_loader, criterion, optimizer, device)\n    print(f\"Train Loss: {train_loss:.4f}, Train Acc: {train_acc:.4f}\")","metadata":{"execution":{"iopub.status.busy":"2024-10-08T06:03:10.721840Z","iopub.execute_input":"2024-10-08T06:03:10.722250Z","iopub.status.idle":"2024-10-08T06:31:23.014685Z","shell.execute_reply.started":"2024-10-08T06:03:10.722201Z","shell.execute_reply":"2024-10-08T06:31:23.013425Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Save the model","metadata":{}},{"cell_type":"code","source":"torch.save(model.state_dict(), 'landmark_model.pth')\nprint(\"Model saved as: landmark_model.pth\")","metadata":{"execution":{"iopub.status.busy":"2024-10-08T06:31:23.016335Z","iopub.execute_input":"2024-10-08T06:31:23.016727Z","iopub.status.idle":"2024-10-08T06:31:24.863284Z","shell.execute_reply.started":"2024-10-08T06:31:23.016689Z","shell.execute_reply":"2024-10-08T06:31:24.862264Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Make pred","metadata":{}},{"cell_type":"code","source":"test_dir = '/kaggle/input/landmark-recognition-2021/test'\ntest_images = []\nfor root, dirs, files in os.walk(test_dir):\n    for file in files:\n        if file.endswith('.jpg'):\n            test_images.append(os.path.join(root, file))\n\ntest_dataset = LandmarkDataset(test_images, transform=transform)\ntest_loader = DataLoader(test_dataset, batch_size=256, shuffle=False, num_workers=4, pin_memory=True)","metadata":{"execution":{"iopub.status.busy":"2024-10-08T06:31:24.864709Z","iopub.execute_input":"2024-10-08T06:31:24.865107Z","iopub.status.idle":"2024-10-08T06:31:29.247370Z","shell.execute_reply.started":"2024-10-08T06:31:24.865071Z","shell.execute_reply":"2024-10-08T06:31:29.246243Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model.eval()\npredictions = []\nids = []\n\nwith torch.no_grad():\n    for inputs in tqdm(test_loader, desc=\"Predicting\"):\n        inputs = inputs.to(device)\n        outputs = model(inputs)\n        probs = torch.softmax(outputs, dim=1)\n        confidence, predicted = probs.max(1)\n        predictions.extend(zip(predicted.cpu().numpy(), confidence.cpu().numpy()))\n        ids.extend([os.path.basename(img_path).split('.')[0] for img_path in test_images[len(ids):len(ids) + len(inputs)]])","metadata":{"execution":{"iopub.status.busy":"2024-10-08T06:31:29.248754Z","iopub.execute_input":"2024-10-08T06:31:29.249134Z","iopub.status.idle":"2024-10-08T06:31:53.979117Z","shell.execute_reply.started":"2024-10-08T06:31:29.249096Z","shell.execute_reply":"2024-10-08T06:31:53.978055Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Create submission file","metadata":{}},{"cell_type":"code","source":"submission_df = pd.DataFrame({\n    'id': ids,\n    'landmarks': [f\"{inv_label_mapping[label]} {conf:.4f}\" for label, conf in predictions]\n})\n\nsubmission_df.to_csv('submission.csv', index=False)\nprint(\"Submission file created: submission.csv\")","metadata":{"execution":{"iopub.status.busy":"2024-10-08T06:31:53.980525Z","iopub.execute_input":"2024-10-08T06:31:53.980898Z","iopub.status.idle":"2024-10-08T06:31:54.036573Z","shell.execute_reply.started":"2024-10-08T06:31:53.980859Z","shell.execute_reply":"2024-10-08T06:31:54.035793Z"},"trusted":true},"execution_count":null,"outputs":[]}]}