train.py

import os
import tensorflow as tf
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras.applications import MobileNetV2
from tensorflow.keras import layers, models
from tensorflow.keras.optimizers import Adam

# Updated path
base_dir = 'data/chest_xray'
train_dir = os.path.join(base_dir, 'train')
val_dir = os.path.join(base_dir, 'val')

# Parameters
IMG_SIZE = (224, 224)
BATCH_SIZE = 32
EPOCHS = 5  # You can increase later

# Data generators
train_gen = ImageDataGenerator(
    rescale=1./255,
    rotation_range=10,
    zoom_range=0.1,
    horizontal_flip=True
)

val_gen = ImageDataGenerator(rescale=1./255)

train_data = train_gen.flow_from_directory(
    train_dir,
    target_size=IMG_SIZE,
    batch_size=BATCH_SIZE,
    class_mode='binary'
)

val_data = val_gen.flow_from_directory(
    val_dir,
    target_size=IMG_SIZE,
    batch_size=BATCH_SIZE,
    class_mode='binary'
)

# MobileNetV2 base
base_model = MobileNetV2(input_shape=(224, 224, 3), include_top=False, weights='imagenet')
base_model.trainable = False

# Custom head
model = models.Sequential([
    base_model,
    layers.GlobalAveragePooling2D(),
    layers.Dense(128, activation='relu'),
    layers.Dropout(0.3),
    layers.Dense(1, activation='sigmoid')  # Binary classifier
])

model.compile(optimizer=Adam(learning_rate=0.0001),
              loss='binary_crossentropy',
              metrics=['accuracy'])

# Train
model.fit(train_data, validation_data=val_data, epochs=EPOCHS)

# Save model
os.makedirs("model", exist_ok=True)
model.save("model/pneumonia_model.h5")
print("✅ Model saved as model/pneumonia_model.h5")