{"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":"gpu","dataSources":[{"sourceId":112509,"databundleVersionId":14254895,"sourceType":"competition"},{"sourceId":14113251,"sourceType":"datasetVersion","datasetId":8990263},{"sourceId":14145231,"sourceType":"datasetVersion","datasetId":9004317},{"sourceId":289705324,"sourceType":"kernelVersion"}],"dockerImageVersionId":31193,"isInternetEnabled":false,"language":"python","sourceType":"notebook","isGpuEnabled":true}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"
\n
\n

🚀 CSIRO Inference: Pure Vision SOTA benchmark

\n

ViT-Base/Small + Tiling + TTA (No Metadata Leakage)

\n
\n
","metadata":{}},{"cell_type":"code","source":"import sys\nimport os\nimport gc\nimport numpy as np\nimport pandas as pd\nfrom pathlib import Path\nfrom tqdm.auto import tqdm\nfrom PIL import Image\nimport cv2\nimport re\n\n# Deep Learning\nimport torch\nimport torch.nn as nn\nimport torch.nn.functional as F\nfrom torch.utils.data import Dataset, DataLoader\nimport pytorch_lightning as pl\nimport albumentations as A\nfrom albumentations.pytorch import ToTensorV2\nimport timm\nfrom sklearn.preprocessing import RobustScaler, MinMaxScaler\n\n# --- CONFIG ---\nclass CFG:\n # ⚠️ IMPORTANT: Update this path to where you uploaded your .ckpt files\n MODEL_DIR = \"/kaggle/input/csiro-nb/models/\" \n \n # Competition Data Paths\n TEST_CSV = \"/kaggle/input/csiro-biomass/test.csv\"\n TEST_DIR = \"/kaggle/input/csiro-biomass/test\"\n TRAIN_CSV = \"/kaggle/input/csiro-biomass/train.csv\"\n \n # Model Architecture\n MODEL_NAME = \"vit_small_patch14_dinov2.lvd142m\"\n \n # Tiling Settings\n TILE_SIZE = 518 \n IMG_LOAD_SIZE = 1036\n \n BATCH_SIZE = 8\n NUM_WORKERS = 2\n \n # Targets\n TARGET_COLS = ['Dry_Total_g', 'GDM_g', 'Dry_Green_g', 'Dry_Dead_g', 'Dry_Clover_g']\n AUX_COLS = ['Height_Ave_cm', 'Pre_GSHH_NDVI'] ","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true,"execution":{"iopub.status.busy":"2026-01-02T21:50:09.789943Z","iopub.execute_input":"2026-01-02T21:50:09.790252Z","iopub.status.idle":"2026-01-02T21:50:09.796495Z","shell.execute_reply.started":"2026-01-02T21:50:09.790229Z","shell.execute_reply":"2026-01-02T21:50:09.795792Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## 🧬 Model Classes","metadata":{}},{"cell_type":"code","source":"class TiledDINO(pl.LightningModule):\n def __init__(self):\n super().__init__()\n # Load Backbone (Pretrained=False for Offline)\n self.backbone = timm.create_model(\n CFG.MODEL_NAME, \n pretrained=False, \n num_classes=0,\n img_size=CFG.TILE_SIZE \n )\n \n embed_dim = self.backbone.num_features\n \n # Heads\n self.head_main = nn.Sequential(\n nn.Linear(embed_dim, 512),\n nn.BatchNorm1d(512),\n nn.SiLU(),\n nn.Dropout(0.1),\n nn.Linear(512, 5)\n )\n \n self.head_aux = nn.Sequential(\n nn.Linear(embed_dim, 256), \n nn.BatchNorm1d(256), \n nn.SiLU(), \n nn.Linear(256, 2) \n )\n def forward(self, x):\n # x: [Batch, 4, 3, 518, 518]\n b, n_tiles, c, h, w = x.shape\n x_flat = x.view(b * n_tiles, c, h, w)\n \n feats = self.backbone(x_flat) \n feats = feats.view(b, n_tiles, -1)\n global_feat = torch.mean(feats, dim=1) \n \n return self.head_main(global_feat), self.head_aux(global_feat)\n\ndef predict_tta(model_list, img_batch, meta_batch):\n # Pass meta_batch to the forward function combining the predictions across all models (ensemble)\n return np.mean([m(img_batch, meta_batch)[0].cpu().numpy() for m in model_list], axis=0)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-01-02T21:50:11.789658Z","iopub.execute_input":"2026-01-02T21:50:11.790405Z","iopub.status.idle":"2026-01-02T21:50:11.796970Z","shell.execute_reply.started":"2026-01-02T21:50:11.790378Z","shell.execute_reply":"2026-01-02T21:50:11.796261Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## 📂 Data Handling & Processing","metadata":{}},{"cell_type":"code","source":"class TiledDataset(Dataset):\n def __init__(self, df, transforms=None):\n self.df = df.reset_index(drop=True)\n self.paths = df['full_path'].values\n self.transforms = transforms\n\n def __len__(self):\n return len(self.df)\n \n def tile_image(self, image):\n H, W = CFG.IMG_LOAD_SIZE, CFG.IMG_LOAD_SIZE\n image = cv2.resize(image, (W, H))\n \n t_sz = CFG.TILE_SIZE\n tiles = []\n for i in range(2):\n for j in range(2):\n tile = image[i*t_sz:(i+1)*t_sz, j*t_sz:(j+1)*t_sz, :]\n tiles.append(tile)\n return tiles\n\n def __getitem__(self, idx):\n path = self.paths[idx]\n try:\n image = np.array(Image.open(path).convert(\"RGB\"))\n except:\n image = np.zeros((CFG.IMG_LOAD_SIZE, CFG.IMG_LOAD_SIZE, 3), dtype=np.uint8)\n \n tiles = self.tile_image(image)\n \n processed_tiles = []\n for t in tiles:\n if self.transforms:\n t = self.transforms(image=t)['image']\n processed_tiles.append(t)\n \n img_tensor = torch.stack(processed_tiles)\n return img_tensor\n\ndef get_transforms():\n return A.Compose([\n A.Resize(CFG.TILE_SIZE, CFG.TILE_SIZE),\n A.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),\n ToTensorV2(),\n ])\n\ndef get_processor():\n print(\"Fitting Scaler on Training Data...\")\n train_df = pd.read_csv(CFG.TRAIN_CSV)\n train_df['image_id'] = train_df['sample_id'].str.split('__').str[0]\n \n # Pivot logic to match training\n meta_cols = ['image_path', 'Sampling_Date']\n df_wide = train_df.pivot_table(\n index=['image_id'] + meta_cols,\n columns='target_name',\n values='target',\n aggfunc='first'\n ).reset_index()\n \n main_scaler = RobustScaler()\n main_scaler.fit(df_wide[CFG.TARGET_COLS])\n print(\"✅ Scaler fitted.\")\n return main_scaler\n\ndef predict_tta(model_list, img_batch):\n return np.mean([m(img_batch)[0].cpu().numpy() for m in model_list], axis=0)","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-01-02T21:50:12.049571Z","iopub.execute_input":"2026-01-02T21:50:12.050191Z","iopub.status.idle":"2026-01-02T21:50:12.059169Z","shell.execute_reply.started":"2026-01-02T21:50:12.050170Z","shell.execute_reply":"2026-01-02T21:50:12.058635Z"}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## 🚀 Main Inference Pipeline\nIncludes **Metadata Encoding** setup and **TTA Loop**.","metadata":{}},{"cell_type":"code","source":"print(\"🚀 Starting Pure Vision Inference...\")\n\npattern = r\"vit_small_patch14_dinov2\\.lvd142m/fold_\\d/.*\\.ckpt$\"\n\n# 1. Load Test Data\ntest_df = pd.read_csv(CFG.TEST_CSV)\ntest_df['image_id'] = test_df['sample_id'].str.split('__').str[0]\n\n# Detect Extension\ntry:\n sample_file = [f for f in os.listdir(CFG.TEST_DIR) if not f.startswith('.')][0]\n ext = os.path.splitext(sample_file)[1]\nexcept:\n ext = \".png\"\n\ntest_df['full_path'] = test_df['image_id'].apply(lambda x: os.path.join(CFG.TEST_DIR, f\"{x}{ext}\"))\n\n# Unique images only\ndf_unique = test_df[['image_id', 'full_path']].drop_duplicates(subset=['image_id']).reset_index(drop=True)\n\ntest_ds = TiledDataset(df_unique, get_transforms())\nloader = DataLoader(test_ds, batch_size=CFG.BATCH_SIZE, shuffle=False, num_workers=CFG.NUM_WORKERS)\n\n# 2. Load Models\nmodels = []\nckpt_files = [\n str(p) for p in Path(CFG.MODEL_DIR).rglob(\"*.ckpt\") \n if re.search(pattern, str(p))\n]\nprint(f\"📂 Found {len(ckpt_files)} checkpoints.\")\n\nfor p in ckpt_files:\n try:\n model = TiledDINO.load_from_checkpoint(str(p))\n model.eval().cuda()\n models.append(model)\n print(f\" ✅ Loaded {p}\")\n except Exception as e:\n print(f\" ⚠️ Skipping {p}: {e}\")\n\nif not models:\n raise RuntimeError(\"No models loaded! Check CFG.MODEL_DIR\")\n\n# 3. Inference with TTA\nprocessor = get_processor()\nall_preds = []\n\nprint(\"🔥 Running Inference (x3 TTA)...\")\n\nwith torch.no_grad():\n for i, img in tqdm(enumerate(loader), desc=\"Predicting\"):\n img = img.cuda()\n\n # TTA 1: Original\n p1 = predict_tta(models, img)\n \n # TTA 2: Horizontal Flip\n img_h = torch.flip(img, dims=[4]) \n p2 = predict_tta(models, img_h)\n \n # TTA 3: Vertical Flip\n img_v = torch.flip(img, dims=[3]) \n p3 = predict_tta(models, img_v)\n \n # TTA 4: Rotate 90 degrees\n img_r1 = torch.rot90(img, 1, [3, 4])\n p4 = predict_tta(models, img_r1)\n \n # TTA 5: Rotate 270 degrees\n img_r3 = torch.rot90(img, 3, [3, 4])\n p5 = predict_tta(models, img_r3)\n\n # Average all 5 TTA predictions\n batch_preds = (p1 + p2 + p3 + p4 + p5) / 5.0\n all_preds.append(batch_preds)\n\n# 4. Post-Processing & Submission\nall_preds = np.vstack(all_preds)\nall_preds = processor.inverse_transform(all_preds) # Scale back to grams\nall_preds = np.maximum(0, all_preds) # Clip negatives\n\nprint(\"Writing Submission file...\")\npred_map = {img_id: preds for img_id, preds in zip(df_unique['image_id'].values, all_preds)}\ntarget_idx = {name: i for i, name in enumerate(CFG.TARGET_COLS)}\n\nfinal_values = []\nsub_template = pd.read_csv(CFG.TEST_CSV)\n\nfor _, row in sub_template.iterrows():\n img_id = row['sample_id'].split('__')[0]\n t_name = row['target_name']\n \n if img_id in pred_map:\n val = pred_map[img_id][target_idx[t_name]]\n final_values.append(val)\n else:\n final_values.append(0.0)\n\nsub_template['target'] = final_values\nsub_template[['sample_id', 'target']].to_csv(\"submission.csv\", index=False)\nprint(\"🎉 Done! submission.csv created successfully.\")","metadata":{"trusted":true,"execution":{"iopub.status.busy":"2026-01-02T21:50:16.078538Z","iopub.execute_input":"2026-01-02T21:50:16.079269Z","iopub.status.idle":"2026-01-02T21:50:20.605009Z","shell.execute_reply.started":"2026-01-02T21:50:16.079241Z","shell.execute_reply":"2026-01-02T21:50:20.604290Z"}},"outputs":[],"execution_count":null}]}