{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.11.13","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"nvidiaTeslaT4","dataSources":[{"sourceId":112509,"databundleVersionId":14254895,"isSourceIdPinned":false,"sourceType":"competition"},{"sourceId":272090986,"sourceType":"kernelVersion"}],"dockerImageVersionId":31154,"isInternetEnabled":false,"language":"python","sourceType":"notebook","isGpuEnabled":true}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"---\n\n# Methodology: CSIRO Pasture Biomass Prediction\n\n## 1. Core Strategy: Predicting Key Components\n\nThe primary goal is to predict five biomass targets. Based on exploratory data analysis (EDA), we identified linear dependencies:\n* `Dry_Total_g` $\\approx$ `Dry_Green_g` + `Dry_Dead_g` + `Dry_Clover_g`\n* `GDM_g` $\\approx$ `Dry_Green_g` + `Dry_Clover_g`\n\nTo avoid redundancy, the model is trained to predict only the **three most visually distinct and/or highest-weighted targets**:\n* `Dry_Total_g` (50% of the score)\n* `GDM_g` (20% of the score)\n* `Dry_Green_g` (10% of the score)\n\nThe remaining two targets (`Dry_Dead_g` and `Dry_Clover_g`) are then **calculated during validation and inference** using subtraction (e.g., `pred_Clover = max(0, pred_GDM - pred_Green)`).\n\n---\n\n## 2. Data Handling & K-Fold Strategy\n\n* **Image Input:** All source images are high-resolution (`2000x1000` pixels).\n* **Two-Stream Processing:** To preserve fine-grained details (like clover leaves) that would be lost by resizing the entire image, the `Dataset` class crops each image into two `1000x1000` patches (a \"left\" and \"right\" half).\n* **High-Resolution Input:** Each `1000x1000` patch is then resized to **`768x768`**, maintaining a high level of detail.\n* **K-Fold Strategy:** We use a **5-Fold Cross-Validation** strategy due to the small dataset (357 images).\n* **Robust Splitting (GroupKFold):** To prevent data leakage (where similar images from the same day are in both train and validation), we use `GroupKFold` grouped by `Sampling_Date`. This ensures the model is validated on dates it has never seen.\n\n---\n\n## 3. Model Architecture: Two-Stream, Multi-Head\n\nThe model uses a \"Two-Stream, Multi-Head\" architecture.\n* **Shared Backbone:** A single `timm` backbone (e.g., `convnext_tiny`) with pre-trained ImageNet weights is used.\n* **Two-Stream Input:**\n    * `img_left` $\\rightarrow$ `backbone` $\\rightarrow$ `features_left`\n    * `img_right` $\\rightarrow$ (same) `backbone` $\\rightarrow$ `features_right`\n* **Fusion:** The two feature vectors are concatenated: `combined_features = torch.cat([features_left, features_right])`.\n* **Multi-Head Output:** This combined vector is fed into **three separate, specialized MLP heads** (one for each target: `head_total`, `head_gdm`, `head_green`) to allow for task specialization.\n\n---\n\n## 4. Data Augmentation\n\nTo compensate for the small dataset, augmentations are applied **independently** to the `img_left` and `img_right` patches.\n* `HorizontalFlip (p=0.5)`\n* `VerticalFlip (p=0.5)`\n* `RandomRotate90 (p=0.5)` (Only 90-degree rotations)\n* `ColorJitter`\n\nThis independent application creates a much larger variety of training combinations.\n\n---\n\n## 5. Loss Function: Weighted SmoothL1Loss\n\nThe model is optimized using a custom weighted loss function that aligns with the competition's scoring metric.\n* **Base Loss:** `nn.SmoothL1Loss` (Huber Loss) is used instead of `MSELoss` to make training more stable and less sensitive to outliers.\n* **Weighted Sum:** The final loss is a weighted sum of the individual losses, using the competition's scoring weights:\n    $$Loss = (0.5 \\cdot Loss_{Total}) + (0.2 \\cdot Loss_{GDM}) + (0.1 \\cdot Loss_{Green})$$\n\n---\n\n## 6. Training Strategy: Two-Stage Fine-Tuning\n\nA two-stage \"Freeze/Unfreeze\" strategy is used to stabilize training on the small dataset.\n* **Stage 1 (Freeze):**\n    * **Epochs:** 1-5\n    * **Action:** The entire `backbone` is frozen. Only the three MLP heads are trained.\n    * **LR:** `1e-4`\n* **Stage 2 (Unfreeze/Fine-Tuning):**\n    * **Epochs:** 6-20\n    * **Action:** The `backbone` is \"unfrozen,\" and the entire model is trained.\n    * **LR:** A very low learning rate (`1e-5`) is used to slowly adapt the backbone features.\n* **Model Saving:** A `ModelCheckpoint` saves the model based on the **highest `Score (R^2)`** on the validation set, *not* the lowest loss. This is critical for capturing the model's peak performance (like the `R^2=0.64` spike at Epoch 11) and ignoring the unstable, overfitted epochs.","metadata":{}},{"cell_type":"code","source":"import os\nimport pandas as pd\nimport numpy as np\nimport torch\nimport torch.nn as nn\nfrom torch.utils.data import Dataset, DataLoader\nimport albumentations as A\nfrom albumentations.pytorch import ToTensorV2\nimport timm\nimport cv2\nfrom tqdm import tqdm\nimport gc\n\n# ===============================================================\n# 1. ⚙️ CONFIGURATION (PHẢI GIỐNG HỆT FILE TRAINING)\n# ===============================================================\nclass CFG:\n    # --- Đường dẫn (Paths) ---\n    # (Hãy điều chỉnh các đường dẫn này cho đúng với môi trường của bạn)\n    BASE_PATH = '/kaggle/input/csiro-biomass'\n    TEST_CSV = os.path.join(BASE_PATH, 'test.csv')\n    TEST_IMAGE_DIR = os.path.join(BASE_PATH, 'test')\n    \n    # Thư mục chứa 5 file .pth\n    MODEL_DIR = '/kaggle/input/csiro/' # Giả sử 5 file .pth nằm cùng thư mục\n    SUBMISSION_FILE = 'submission.csv'\n    \n    # --- Cài đặt Mô hình (PHẢI TRÙNG KHỚP) ---\n    MODEL_NAME = 'convnext_tiny' # PHẢI GIỐNG HỆT LÚC TRAIN\n    IMG_SIZE = 768               # PHẢI GIỐNG HỆT LÚC TRAIN\n    \n    # --- Cài đặt Inference ---\n    DEVICE = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n    BATCH_SIZE = 1 # Có thể tăng batch size khi inference\n    NUM_WORKERS = 1\n    N_FOLDS = 5\n    \n    # --- Mục tiêu & Loss (PHẢI TRÙNG KHỚP) ---\n    # 3 mục tiêu model đã dự đoán\n    TARGET_COLS = ['Dry_Total_g', 'GDM_g', 'Dry_Green_g']\n    \n    # 5 mục tiêu để nộp bài\n    ALL_TARGET_COLS = ['Dry_Green_g', 'Dry_Dead_g', 'Dry_Clover_g', 'GDM_g', 'Dry_Total_g']\n\nprint(f\"Sử dụng thiết bị: {CFG.DEVICE}\")\nprint(f\"Backbone mô hình: {CFG.MODEL_NAME}\")\nprint(f\"Kích thước ảnh inference: {CFG.IMG_SIZE}x{CFG.IMG_SIZE}\")\n\n\n# ===============================================================\n# 2. 🏞️ AUGMENTATIONS (CHỈ DÙNG VALIDATION)\n# ===============================================================\nfrom albumentations import (\n    Compose, \n    Resize, \n    Normalize,\n    HorizontalFlip, \n    VerticalFlip\n)\n\ndef get_tta_transforms():\n    \"\"\"\n    Trả về một LIST các pipeline transform cho TTA.\n    Mỗi pipeline là một \"view\" khác nhau của ảnh.\n    \"\"\"\n    \n    # Đây là các bước chuẩn hóa cơ bản\n    base_transforms = [\n        Normalize(\n            mean=[0.485, 0.456, 0.406],\n            std=[0.229, 0.224, 0.225]\n        ),\n        ToTensorV2()\n    ]\n    \n    # -----------------\n    # View 1: Ảnh gốc (Chỉ Resize + Normalize)\n    # -----------------\n    original_view = Compose([\n        Resize(CFG.IMG_SIZE, CFG.IMG_SIZE),\n        *base_transforms\n    ])\n    \n    # -----------------\n    # View 2: Lật ngang (HFlip)\n    # -----------------\n    hflip_view = Compose([\n        HorizontalFlip(p=1.0), # Luôn luôn lật\n        Resize(CFG.IMG_SIZE, CFG.IMG_SIZE),\n        *base_transforms\n    ])\n    \n    # -----------------\n    # View 3: Lật dọc (VFlip)\n    # -----------------\n    vflip_view = Compose([\n        VerticalFlip(p=1.0), # Luôn luôn lật\n        Resize(CFG.IMG_SIZE, CFG.IMG_SIZE),\n        *base_transforms\n    ])\n    \n    return [original_view, hflip_view, vflip_view]\n\nprint(\"Đã định nghĩa hàm get_tta_transforms().\")\n\n\nclass TestBiomassDataset(Dataset):\n    \"\"\"\n    Dataset tùy chỉnh cho ảnh test (Chiến lược \"Hai luồng\").\n    Sửa đổi để chấp nhận một pipeline transform cụ thể cho TTA.\n    \"\"\"\n    def __init__(self, df, transform_pipeline, image_dir):\n        self.df = df\n        # (SỬA ĐỔI) Chấp nhận một pipeline đã được khởi tạo\n        self.transforms = transform_pipeline \n        self.image_dir = image_dir\n        self.image_paths = df['image_path'].values\n\n    def __len__(self):\n        return len(self.df)\n\n    def __getitem__(self, idx):\n        # 1. Lấy thông tin\n        img_path_suffix = self.image_paths[idx]\n        \n        # 2. Đọc ảnh gốc (2000x1000)\n        filename = os.path.basename(img_path_suffix)\n        full_path = os.path.join(self.image_dir, filename)\n        \n        image = cv2.imread(full_path)\n        if image is None:\n            print(f\"Warning: Không thể đọc ảnh: {full_path}. Trả về ảnh đen.\")\n            image = np.zeros((1000, 2000, 3), dtype=np.uint8)\n        \n        image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)\n        \n        # 3. Cắt (Crop) thành 2 ảnh (Trái và Phải)\n        height, width, _ = image.shape\n        mid_point = width // 2\n        img_left = image[:, :mid_point]\n        img_right = image[:, mid_point:]\n        \n        # 4. Áp dụng TTA Transform (CÙNG MỘT TRANSFORM cho cả 2)\n        # (Ví dụ: Cả 2 ảnh đều bị lật ngang)\n        img_left_tensor = self.transforms(image=img_left)['image']\n        img_right_tensor = self.transforms(image=img_right)['image']\n        \n        # 5. Trả về\n        return img_left_tensor, img_right_tensor\n\n# ===============================================================\n# 4. 🧠 MODEL ARCHITECTURE (SAO CHÉP TỪ FILE TRAIN)\n# ===============================================================\nclass BiomassModel(nn.Module):\n    \"\"\"\n    Kiến trúc mô hình (Hai luồng, Ba đầu ra)\n    PHẢI GIỐNG HỆT file training.\n    \"\"\"\n    def __init__(self, model_name, pretrained, n_targets=3):\n        super(BiomassModel, self).__init__()\n        \n        self.backbone = timm.create_model(\n            model_name,\n            pretrained=pretrained, # Sẽ là False khi inference\n            num_classes=0,\n            global_pool='avg'\n        )\n        \n        self.n_features = self.backbone.num_features\n        self.n_combined_features = self.n_features * 2\n        \n        # --- Đầu cho Dry_Total_g ---\n        self.head_total = nn.Sequential(\n            nn.Linear(self.n_combined_features, self.n_combined_features // 2),\n            nn.ReLU(),\n            nn.Dropout(0.3),\n            nn.Linear(self.n_combined_features // 2, 1)\n        )\n        \n        # --- Đầu cho GDM_g ---\n        self.head_gdm = nn.Sequential(\n            nn.Linear(self.n_combined_features, self.n_combined_features // 2),\n            nn.ReLU(),\n            nn.Dropout(0.3),\n            nn.Linear(self.n_combined_features // 2, 1)\n        )\n        \n        # --- Đầu cho Dry_Green_g ---\n        self.head_green = nn.Sequential(\n            nn.Linear(self.n_combined_features, self.n_combined_features // 2),\n            nn.ReLU(),\n            nn.Dropout(0.3),\n            nn.Linear(self.n_combined_features // 2, 1)\n        )\n\n    def forward(self, img_left, img_right):\n        features_left = self.backbone(img_left)\n        features_right = self.backbone(img_right)\n        combined = torch.cat([features_left, features_right], dim=1)\n        \n        out_total = self.head_total(combined)\n        out_gdm = self.head_gdm(combined)\n        out_green = self.head_green(combined)\n        \n        return out_total, out_gdm, out_green\n\n\ndef predict_one_view(models_list, test_loader, device):\n    \"\"\"\n    Hàm con: Chạy dự đoán ensemble 5-fold cho MỘT view TTA.\n    \"\"\"\n    view_preds_3 = {'total': [], 'gdm': [], 'green': []}\n    \n    with torch.no_grad():\n        for (img_left, img_right) in tqdm(test_loader, desc=\"  Predicting View\", leave=False):\n            img_left = img_left.to(device)\n            img_right = img_right.to(device)\n            \n            batch_preds_3_folds = {'total': [], 'gdm': [], 'green': []}\n            \n            # 1. Vòng lặp Ensemble 5-Fold\n            for model in models_list:\n                pred_total, pred_gdm, pred_green = model(img_left, img_right)\n                batch_preds_3_folds['total'].append(pred_total.cpu())\n                batch_preds_3_folds['gdm'].append(pred_gdm.cpu())\n                batch_preds_3_folds['green'].append(pred_green.cpu())\n            \n            # 2. Lấy trung bình 5 Fold\n            avg_pred_total = torch.mean(torch.stack(batch_preds_3_folds['total']), dim=0)\n            avg_pred_gdm = torch.mean(torch.stack(batch_preds_3_folds['gdm']), dim=0)\n            avg_pred_green = torch.mean(torch.stack(batch_preds_3_folds['green']), dim=0)\n            \n            view_preds_3['total'].append(avg_pred_total.numpy())\n            view_preds_3['gdm'].append(avg_pred_gdm.numpy())\n            view_preds_3['green'].append(avg_pred_green.numpy())\n\n    # 3. Ghép kết quả các batch của view này\n    preds_np = {\n        'total': np.concatenate(view_preds_3['total']).flatten(),\n        'gdm':   np.concatenate(view_preds_3['gdm']).flatten(),\n        'green': np.concatenate(view_preds_3['green']).flatten()\n    }\n    return preds_np\n\n\ndef run_inference_with_tta():\n    \"\"\"\n    Hàm inference chính, thực hiện TTA x Ensemble.\n    \"\"\"\n    print(f\"\\n{'='*50}\")\n    print(f\"🚀 BẮT ĐẦU INFERENCE (với TTA) 🚀\")\n    print(f\"{'='*50}\")\n\n    # --- 1. Tải Dữ liệu Test ---\n    print(f\"Đang tải {CFG.TEST_CSV}...\")\n    try:\n        test_df_long = pd.read_csv(CFG.TEST_CSV)\n        test_df_unique = test_df_long.drop_duplicates(subset=['image_path']).reset_index(drop=True)\n        print(f\"Tìm thấy {len(test_df_unique)} ảnh test duy nhất.\")\n    except FileNotFoundError:\n        print(f\"LỖI: Không tìm thấy {CFG.TEST_CSV}\")\n        return None, None, None\n\n    # --- 2. Tải 5 Mô hình (Ensemble) ---\n    print(\"\\nĐang tải 5 mô hình đã huấn luyện...\")\n    models_list = []\n    # (Code tải 5 mô hình... giống hệt bước 16 của file trước)\n    for fold in range(CFG.N_FOLDS):\n        model_path = os.path.join(CFG.MODEL_DIR, f'best_model_fold{fold}.pth')\n        if not os.path.exists(model_path):\n            print(f\"LỖI: Không tìm thấy file mô hình: {model_path}\")\n            return None, None, None\n        model = BiomassModel(CFG.MODEL_NAME, pretrained=False)\n        try:\n            model.load_state_dict(torch.load(model_path, map_location=CFG.DEVICE))\n        except RuntimeError:\n            state_dict = torch.load(model_path, map_location=CFG.DEVICE)\n            from collections import OrderedDict\n            new_state_dict = OrderedDict()\n            for k, v in state_dict.items():\n                name = k.replace('module.', '')\n                new_state_dict[name] = v\n            model.load_state_dict(new_state_dict)\n        model.eval()\n        model.to(CFG.DEVICE)\n        models_list.append(model)\n    print(f\"✓ Đã tải thành công {len(models_list)} mô hình.\")\n\n    # --- 3. Vòng lặp TTA (Vòng lặp ngoài) ---\n    tta_transforms = get_tta_transforms()\n    print(f\"\\nBắt đầu dự đoán với {len(tta_transforms)} TTA views...\")\n    \n    all_tta_view_preds = [] # List để lưu kết quả của mỗi view TTA\n\n    for i, tta_transform in enumerate(tta_transforms):\n        print(f\"--- Đang chạy TTA View {i+1}/{len(tta_transforms)} ---\")\n        \n        # Tạo Dataset/Loader MỚI cho view TTA này\n        test_dataset = TestBiomassDataset(\n            df=test_df_unique,\n            transform_pipeline=tta_transform, # Truyền pipeline TTA\n            image_dir=CFG.TEST_IMAGE_DIR\n        )\n        test_loader = DataLoader(\n            test_dataset,\n            batch_size=CFG.BATCH_SIZE,\n            shuffle=False,\n            num_workers=CFG.NUM_WORKERS,\n            pin_memory=True\n        )\n        \n        # Chạy ensemble 5-fold cho view này\n        view_preds_np = predict_one_view(models_list, test_loader, CFG.DEVICE)\n        all_tta_view_preds.append(view_preds_np)\n        print(f\"✓ Hoàn thành TTA View {i+1}\")\n\n    # --- 4. Ensemble (Lấy trung bình) kết quả TTA ---\n    print(\"\\nĐang ensemble kết quả của các TTA views...\")\n    final_ensembled_preds = {\n        'total': np.mean([d['total'] for d in all_tta_view_preds], axis=0),\n        'gdm':   np.mean([d['gdm'] for d in all_tta_view_preds], axis=0),\n        'green': np.mean([d['green'] for d in all_tta_view_preds], axis=0)\n    }\n    \n    print(\"✓ Dự đoán hoàn tất.\")\n    \n    del models_list, test_loader, test_dataset\n    gc.collect()\n    torch.cuda.empty_cache()\n    \n    return final_ensembled_preds, test_df_long, test_df_unique\n# ===============================================================\n# 6. ✍️ HÀM TẠO FILE SUBMISSION\n# ===============================================================\ndef create_submission(preds_np, test_df_long, test_df_unique):\n    \"\"\"\n    Hàm này nhận 3 dự đoán đã ensemble,\n    tính toán 2 dự đoán còn lại,\n    và định dạng file nộp bài.\n    \"\"\"\n    if preds_np is None:\n        print(\"Bỏ qua tạo submission do lỗi ở trên.\")\n        return\n\n    print(\"\\nĐang hậu xử lý và tạo file submission...\")\n\n    # 1. Lấy 3 dự đoán đã ensemble\n    pred_total_final = preds_np['total']\n    pred_gdm_final = preds_np['gdm']\n    pred_green_final = preds_np['green']\n\n    # 2. Tính 2 mục tiêu còn lại (Hậu xử lý)\n    # Dùng np.maximum(0, ...) để đảm bảo không có giá trị âm\n    pred_clover_final = np.maximum(0, pred_gdm_final - pred_green_final)\n    pred_dead_final = np.maximum(0, pred_total_final - pred_gdm_final)\n\n    # 3. Tạo một DataFrame \"wide\" chứa 5 dự đoán\n    # (Đảm bảo thứ tự 5 cột giống CFG.ALL_TARGET_COLS)\n    preds_wide_df = pd.DataFrame({\n        'image_path': test_df_unique['image_path'],\n        'Dry_Green_g': pred_green_final,\n        'Dry_Dead_g': pred_dead_final,\n        'Dry_Clover_g': pred_clover_final,\n        'GDM_g': pred_gdm_final,\n        'Dry_Total_g': pred_total_final\n    })\n\n    # 4. \"Un-pivot\" DataFrame (Chuyển sang dạng \"long\")\n    # Biến nó từ 5 cột về dạng \"long\" (giống sample_submission)\n    preds_long_df = preds_wide_df.melt(\n        id_vars=['image_path'],\n        value_vars=CFG.ALL_TARGET_COLS, # 5 cột mục tiêu\n        var_name='target_name',        # Cột tên mục tiêu\n        value_name='target'            # Cột giá trị dự đoán\n    )\n\n    # 5. Merge với file test.csv gốc (test_df_long)\n    # Đây là bước quan trọng để lấy đúng 'sample_id'\n    # (ví dụ: 'ID1001187975__Dry_Clover_g')\n    submission_df = pd.merge(\n        test_df_long[['sample_id', 'image_path', 'target_name']],\n        preds_long_df,\n        on=['image_path', 'target_name'],\n        how='left'\n    )\n\n    # 6. Dọn dẹp và Lưu\n    # Chỉ giữ lại 2 cột được yêu cầu\n    submission_df = submission_df[['sample_id', 'target']]\n    \n    # Lưu file\n    submission_df.to_csv(CFG.SUBMISSION_FILE, index=False)\n\n    print(f\"\\n🎉 HOÀN TẤT! Đã lưu file submission tại: {CFG.SUBMISSION_FILE}\")\n    print(\"--- 5 hàng đầu của file submission ---\")\n    print(submission_df.head())\n    print(\"\\n--- 5 hàng cuối của file submission ---\")\n    print(submission_df.tail())\n    \n# ===============================================================\n# 8. 🏁 CHẠY CHƯƠNG TRÌNH (Đã sửa)\n# ===============================================================\nif __name__ == \"__main__\":\n    # 1. Chạy dự đoán (đã bao gồm TTA)\n    all_preds_np, df_long, df_unique = run_inference_with_tta()\n    \n    # 2. Tạo file submission (Hàm create_submission giữ nguyên)\n    create_submission(all_preds_np, df_long, df_unique)","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true,"execution":{"iopub.status.busy":"2025-10-30T13:37:31.585364Z","iopub.execute_input":"2025-10-30T13:37:31.586001Z","iopub.status.idle":"2025-10-30T13:37:37.021319Z","shell.execute_reply.started":"2025-10-30T13:37:31.585975Z","shell.execute_reply":"2025-10-30T13:37:37.020506Z"}},"outputs":[],"execution_count":null},{"cell_type":"code","source":"","metadata":{"trusted":true},"outputs":[],"execution_count":null}]}