# %% [code] {"execution":{"iopub.status.busy":"2025-12-17T11:52:23.341580Z","iopub.execute_input":"2025-12-17T11:52:23.341853Z","iopub.status.idle":"2025-12-17T11:52:28.244852Z","shell.execute_reply.started":"2025-12-17T11:52:23.341822Z","shell.execute_reply":"2025-12-17T11:52:28.244035Z"}} # !pip install --force-reinstall protobuf==3.20.3 # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T11:52:28.247073Z","iopub.execute_input":"2025-12-17T11:52:28.247411Z","iopub.status.idle":"2025-12-17T11:52:28.252992Z","shell.execute_reply.started":"2025-12-17T11:52:28.247372Z","shell.execute_reply":"2025-12-17T11:52:28.252262Z"}} import os os.environ['PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION'] = 'python' # %% [markdown] # # Data Loading # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:47.868297Z","iopub.execute_input":"2025-12-17T08:00:47.868515Z","iopub.status.idle":"2025-12-17T08:00:47.901451Z","shell.execute_reply.started":"2025-12-17T08:00:47.868499Z","shell.execute_reply":"2025-12-17T08:00:47.900676Z"}} import numpy as np import pandas as pd # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:47.931403Z","iopub.execute_input":"2025-12-17T08:00:47.931628Z","iopub.status.idle":"2025-12-17T08:00:47.943764Z","shell.execute_reply.started":"2025-12-17T08:00:47.931611Z","shell.execute_reply":"2025-12-17T08:00:47.943208Z"}} train = pd.read_csv('/kaggle/input/csiro-biomass/train.csv') test = pd.read_csv('/kaggle/input/csiro-biomass/test.csv') # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:47.944762Z","iopub.execute_input":"2025-12-17T08:00:47.944962Z","iopub.status.idle":"2025-12-17T08:00:47.955497Z","shell.execute_reply.started":"2025-12-17T08:00:47.944931Z","shell.execute_reply":"2025-12-17T08:00:47.954913Z"}} train.head() # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:47.997313Z","iopub.execute_input":"2025-12-17T08:00:47.997746Z","iopub.status.idle":"2025-12-17T08:00:48.002304Z","shell.execute_reply.started":"2025-12-17T08:00:47.997728Z","shell.execute_reply":"2025-12-17T08:00:48.001708Z"}} train.shape # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:48.003402Z","iopub.execute_input":"2025-12-17T08:00:48.003592Z","iopub.status.idle":"2025-12-17T08:00:48.015864Z","shell.execute_reply.started":"2025-12-17T08:00:48.003578Z","shell.execute_reply":"2025-12-17T08:00:48.015171Z"}} del train['sample_id'] # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:48.016604Z","iopub.execute_input":"2025-12-17T08:00:48.016837Z","iopub.status.idle":"2025-12-17T08:00:48.029766Z","shell.execute_reply.started":"2025-12-17T08:00:48.016818Z","shell.execute_reply":"2025-12-17T08:00:48.029016Z"}} colum = train.columns # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:48.031415Z","iopub.execute_input":"2025-12-17T08:00:48.031652Z","iopub.status.idle":"2025-12-17T08:00:48.044271Z","shell.execute_reply.started":"2025-12-17T08:00:48.031636Z","shell.execute_reply":"2025-12-17T08:00:48.043487Z"}} colum = [ 'Sampling_Date', 'State', 'Species', 'Pre_GSHH_NDVI', 'Height_Ave_cm'] # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:48.045132Z","iopub.execute_input":"2025-12-17T08:00:48.045470Z","iopub.status.idle":"2025-12-17T08:00:48.059781Z","shell.execute_reply.started":"2025-12-17T08:00:48.045449Z","shell.execute_reply":"2025-12-17T08:00:48.059225Z"}} train = train.drop(columns=colum,axis=1) # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:48.060575Z","iopub.execute_input":"2025-12-17T08:00:48.060863Z","iopub.status.idle":"2025-12-17T08:00:48.077152Z","shell.execute_reply.started":"2025-12-17T08:00:48.060840Z","shell.execute_reply":"2025-12-17T08:00:48.076575Z"}} train.head() # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:48.077888Z","iopub.execute_input":"2025-12-17T08:00:48.078348Z","iopub.status.idle":"2025-12-17T08:00:48.095747Z","shell.execute_reply.started":"2025-12-17T08:00:48.078331Z","shell.execute_reply":"2025-12-17T08:00:48.095155Z"}} train = train.pivot_table( index="image_path", columns="target_name", values="target" ).reset_index() # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:48.096508Z","iopub.execute_input":"2025-12-17T08:00:48.096782Z","iopub.status.idle":"2025-12-17T08:00:48.106228Z","shell.execute_reply.started":"2025-12-17T08:00:48.096760Z","shell.execute_reply":"2025-12-17T08:00:48.105441Z"}} train.head() # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:48.106951Z","iopub.execute_input":"2025-12-17T08:00:48.107658Z","iopub.status.idle":"2025-12-17T08:00:48.119257Z","shell.execute_reply.started":"2025-12-17T08:00:48.107635Z","shell.execute_reply":"2025-12-17T08:00:48.118596Z"}} target_cols = [ "Dry_Clover_g", "Dry_Dead_g", "Dry_Green_g", "Dry_Total_g", "GDM_g" ] train = train[["image_path"] + target_cols] # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:48.122851Z","iopub.execute_input":"2025-12-17T08:00:48.123076Z","iopub.status.idle":"2025-12-17T08:00:48.137461Z","shell.execute_reply.started":"2025-12-17T08:00:48.123061Z","shell.execute_reply":"2025-12-17T08:00:48.136720Z"}} train.shape # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:48.138136Z","iopub.execute_input":"2025-12-17T08:00:48.138461Z","iopub.status.idle":"2025-12-17T08:00:48.149185Z","shell.execute_reply.started":"2025-12-17T08:00:48.138444Z","shell.execute_reply":"2025-12-17T08:00:48.148445Z"}} # !pip install protobuf # !pip uninstall protobuf -y # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:48.149829Z","iopub.execute_input":"2025-12-17T08:00:48.150020Z","iopub.status.idle":"2025-12-17T08:00:48.162252Z","shell.execute_reply.started":"2025-12-17T08:00:48.149998Z","shell.execute_reply":"2025-12-17T08:00:48.161630Z"}} # import os # os.environ["PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION"] = "python" # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T11:47:36.549407Z","iopub.execute_input":"2025-12-17T11:47:36.549687Z","iopub.status.idle":"2025-12-17T11:47:36.555363Z","shell.execute_reply.started":"2025-12-17T11:47:36.549660Z","shell.execute_reply":"2025-12-17T11:47:36.554638Z"}} # from tensorflow import keras from keras.preprocessing.image import load_img, img_to_array from keras.models import Sequential from keras.layers import Conv2D, MaxPooling2D, Dense, Dropout, Flatten from sklearn.model_selection import train_test_split # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:48.187754Z","iopub.execute_input":"2025-12-17T08:00:48.188008Z","iopub.status.idle":"2025-12-17T08:00:48.201749Z","shell.execute_reply.started":"2025-12-17T08:00:48.187986Z","shell.execute_reply":"2025-12-17T08:00:48.201161Z"}} train_df, valid_df = train_test_split(train,test_size=0.2,random_state=42) # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:48.202484Z","iopub.execute_input":"2025-12-17T08:00:48.202808Z","iopub.status.idle":"2025-12-17T08:00:48.215710Z","shell.execute_reply.started":"2025-12-17T08:00:48.202791Z","shell.execute_reply":"2025-12-17T08:00:48.215186Z"}} train_df = train_df.reset_index(drop=True) valid_df = valid_df.reset_index(drop=True) # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:48.216308Z","iopub.execute_input":"2025-12-17T08:00:48.216559Z","iopub.status.idle":"2025-12-17T08:00:48.230717Z","shell.execute_reply.started":"2025-12-17T08:00:48.216542Z","shell.execute_reply":"2025-12-17T08:00:48.230019Z"}} IMG_SIZE = (224,224) def load_image(path): img = load_img(path,target_size=IMG_SIZE) img = img_to_array(img)/255.0 return img # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:48.231459Z","iopub.execute_input":"2025-12-17T08:00:48.231695Z","iopub.status.idle":"2025-12-17T08:00:48.244847Z","shell.execute_reply.started":"2025-12-17T08:00:48.231680Z","shell.execute_reply":"2025-12-17T08:00:48.244077Z"}} path = "/kaggle/input/csiro-biomass/" def train_generator(df, batch_size=16): while True: df = df.sample(frac=1).reset_index(drop=True) for i in range(0, len(df), batch_size): batch_df = df.iloc[i:i+batch_size] images = [] targets = [] for _, row in batch_df.iterrows(): images.append(load_image(path + row["image_path"])) targets.append(row[target_cols].values.astype(np.float32)) yield np.array(images, dtype=np.float32), np.array(targets, dtype=np.float32) # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:48.245611Z","iopub.execute_input":"2025-12-17T08:00:48.245851Z","iopub.status.idle":"2025-12-17T08:00:48.257303Z","shell.execute_reply.started":"2025-12-17T08:00:48.245827Z","shell.execute_reply":"2025-12-17T08:00:48.256592Z"}} def val_generator(df, batch_size=16): while True: for i in range(0, len(df), batch_size): batch_df = df.iloc[i:i+batch_size] images = [] targets = [] for _, row in batch_df.iterrows(): images.append(load_image(path + row["image_path"])) targets.append(row[target_cols].values.astype(np.float32)) yield np.array(images, dtype=np.float32), np.array(targets, dtype=np.float32) # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:48.258114Z","iopub.execute_input":"2025-12-17T08:00:48.258340Z","iopub.status.idle":"2025-12-17T08:00:50.153856Z","shell.execute_reply.started":"2025-12-17T08:00:48.258320Z","shell.execute_reply":"2025-12-17T08:00:50.153314Z"}} model = Sequential() model.add(Conv2D(32,kernel_size=(3,3),activation='relu',input_shape=(224,224,3))) model.add(MaxPooling2D(pool_size=(2,2),strides=2,padding='valid')) model.add(Conv2D(64,kernel_size=(3,3),activation='relu')) model.add(MaxPooling2D(pool_size=(2,2),strides=2,padding='valid')) model.add(Conv2D(128,kernel_size=(3,3),activation='relu')) model.add(MaxPooling2D(pool_size=(2,2),strides=2,padding='valid')) model.add(Flatten()) model.add(Dense(32,activation='relu')) model.add(Dropout(0.1)) model.add(Dense(64,activation='relu')) model.add(Dropout(0.1)) model.add(Dense(128,activation='relu')) model.add(Dropout(0.1)) model.add(Dense(5,activation='linear')) # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:50.154647Z","iopub.execute_input":"2025-12-17T08:00:50.154906Z","iopub.status.idle":"2025-12-17T08:00:50.176172Z","shell.execute_reply.started":"2025-12-17T08:00:50.154885Z","shell.execute_reply":"2025-12-17T08:00:50.175638Z"}} model.summary() # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:50.176768Z","iopub.execute_input":"2025-12-17T08:00:50.177032Z","iopub.status.idle":"2025-12-17T08:00:50.189110Z","shell.execute_reply.started":"2025-12-17T08:00:50.177010Z","shell.execute_reply":"2025-12-17T08:00:50.188428Z"}} model.compile(optimizer='adam',loss='mse',metrics=['mae']) # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:50.189852Z","iopub.execute_input":"2025-12-17T08:00:50.190172Z","iopub.status.idle":"2025-12-17T08:00:50.696512Z","shell.execute_reply.started":"2025-12-17T08:00:50.190156Z","shell.execute_reply":"2025-12-17T08:00:50.695895Z"}} BATCH_SIZE = 16 EPOCHS = 10 # model.fit( # train_generator(train_df, BATCH_SIZE), # steps_per_epoch=max(1, len(train_df) // BATCH_SIZE), # validation_data=val_generator(valid_df, BATCH_SIZE), # validation_steps=max(1, len(valid_df) // BATCH_SIZE), # epochs=EPOCHS # ) from keras.models import load_model model = load_model("/kaggle/input/cnn-model1/keras/default/1/cnn-model1.keras") # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:50.697321Z","iopub.execute_input":"2025-12-17T08:00:50.697565Z","iopub.status.idle":"2025-12-17T08:00:50.701122Z","shell.execute_reply.started":"2025-12-17T08:00:50.697544Z","shell.execute_reply":"2025-12-17T08:00:50.700368Z"}} # model.save("cnn-model1.keras") # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:50.701847Z","iopub.execute_input":"2025-12-17T08:00:50.702076Z","iopub.status.idle":"2025-12-17T08:00:50.717984Z","shell.execute_reply.started":"2025-12-17T08:00:50.702049Z","shell.execute_reply":"2025-12-17T08:00:50.717459Z"}} test_images = ( test[["image_path"]] .drop_duplicates() .reset_index(drop=True) ) # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:50.718736Z","iopub.execute_input":"2025-12-17T08:00:50.719579Z","iopub.status.idle":"2025-12-17T08:00:50.732938Z","shell.execute_reply.started":"2025-12-17T08:00:50.719561Z","shell.execute_reply":"2025-12-17T08:00:50.732395Z"}} test_images.head() # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:50.733635Z","iopub.execute_input":"2025-12-17T08:00:50.733868Z","iopub.status.idle":"2025-12-17T08:00:50.746384Z","shell.execute_reply.started":"2025-12-17T08:00:50.733849Z","shell.execute_reply":"2025-12-17T08:00:50.745783Z"}} path # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:50.749120Z","iopub.execute_input":"2025-12-17T08:00:50.749605Z","iopub.status.idle":"2025-12-17T08:00:50.862274Z","shell.execute_reply.started":"2025-12-17T08:00:50.749588Z","shell.execute_reply":"2025-12-17T08:00:50.861691Z"}} X_test = np.array( [load_image(path+p) for p in test_images["image_path"]], dtype=np.float32 ) # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:50.863039Z","iopub.execute_input":"2025-12-17T08:00:50.863306Z","iopub.status.idle":"2025-12-17T08:00:50.867958Z","shell.execute_reply.started":"2025-12-17T08:00:50.863280Z","shell.execute_reply":"2025-12-17T08:00:50.867293Z"}} def test_generator(df, batch_size=8): for i in range(0, len(df), batch_size): batch_df = df.iloc[i:i+batch_size] images = [load_image(path + p) for p in batch_df["image_path"]] yield (np.array(images, dtype=np.float32),) # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:50.868666Z","iopub.execute_input":"2025-12-17T08:00:50.868882Z","iopub.status.idle":"2025-12-17T08:00:52.189441Z","shell.execute_reply.started":"2025-12-17T08:00:50.868867Z","shell.execute_reply":"2025-12-17T08:00:52.188798Z"}} preds = model.predict( test_generator(test_images, BATCH_SIZE), steps=int(np.ceil(len(test_images)/BATCH_SIZE)) ) # %% [code] {"execution":{"iopub.status.busy":"2025-12-17T08:00:52.190467Z","iopub.execute_input":"2025-12-17T08:00:52.190712Z","iopub.status.idle":"2025-12-17T08:00:52.312783Z","shell.execute_reply.started":"2025-12-17T08:00:52.190695Z","shell.execute_reply":"2025-12-17T08:00:52.312187Z"}} # FINAL SUBMISSION CELL - Replace your current one completely import pandas as pd import numpy as np import os # Load sample submission FIRST (defines exact format) sample = pd.read_csv("/kaggle/input/csiro-biomass/sample_submission.csv") print("Sample shape:", sample.shape) print("Sample columns:", sample.columns.tolist()) # Generate predictions ONCE (your generator works) test_preds = model.predict( test_generator(test_images, batch_size=8), steps=int(np.ceil(len(test_images) / 8)), verbose=1 ) # Create predictions DataFrame pred_df = pd.DataFrame(test_preds, columns=target_cols) pred_df['image_path'] = test_images['image_path'].values # Match lengths exactly # Create long format matching sample_submission structure submission = pred_df.melt( id_vars='image_path', value_vars=target_cols, var_name='target_name', value_name='target' ) # Generate sample_id exactly like sample_submission submission['sample_id'] = ( submission['image_path'].str.split('/').str[-1] .str.replace('.jpg', '', regex=False) + '__' + submission['target_name'] ) # Reorder columns to match sample_submission exactly submission = submission[['sample_id', 'target']] # VALIDATION CHECKS assert len(submission) == len(sample), f"Row mismatch: {len(submission)} vs {len(sample)}" assert list(submission.columns) == list(sample.columns), "Column mismatch" assert submission.isna().sum().sum() == 0, "NaNs found!" # Save to EXACT correct path submission.to_csv('/kaggle/working/submission.csv', index=False) # FINAL VERIFICATION print("Files:", os.listdir('/kaggle/working/')) print("Submission shape:", pd.read_csv('/kaggle/working/submission.csv').shape) print("Submission head:") print(pd.read_csv('/kaggle/working/submission.csv').head()) print("✅ Submission ready!")