Experiment - Why computer vision teaches machines to see

Problem:We want a computer to recognize objects in pictures, like how we see and understand images.

Current Metrics:Training accuracy: 98%, Validation accuracy: 70%

Issue:The model is overfitting. It learns the training images too well but does not do well on new images.

Your Task

Reduce overfitting so validation accuracy improves to at least 85%, while keeping training accuracy below 92%.

Keep the same dataset and model architecture (a simple CNN).

Only change training settings or add regularization techniques.

Hint 1

Hint 2

Hint 3

Solution

Computer Vision

import tensorflow as tf
from tensorflow.keras import layers, models
from tensorflow.keras.preprocessing.image import ImageDataGenerator

# Load example dataset (CIFAR-10)
(X_train, y_train), (X_test, y_test) = tf.keras.datasets.cifar10.load_data()

# Normalize pixel values
X_train, X_test = X_train / 255.0, X_test / 255.0

# Data augmentation
datagen = ImageDataGenerator(
    rotation_range=15,
    width_shift_range=0.1,
    height_shift_range=0.1,
    horizontal_flip=True
)
datagen.fit(X_train)

# Build CNN model with dropout
model = models.Sequential([
    layers.Conv2D(32, (3,3), activation='relu', input_shape=(32,32,3)),
    layers.MaxPooling2D((2,2)),
    layers.Dropout(0.25),
    layers.Conv2D(64, (3,3), activation='relu'),
    layers.MaxPooling2D((2,2)),
    layers.Dropout(0.25),
    layers.Flatten(),
    layers.Dense(128, activation='relu'),
    layers.Dropout(0.5),
    layers.Dense(10, activation='softmax')
])

model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])

# Train with data augmentation and early stopping
early_stop = tf.keras.callbacks.EarlyStopping(monitor='val_loss', patience=5, restore_best_weights=True)

history = model.fit(datagen.flow(X_train, y_train, batch_size=64), epochs=50, validation_data=(X_test, y_test), callbacks=[early_stop])

Added dropout layers after convolution and dense layers to reduce overfitting.

Used data augmentation to create more varied training images.

Added early stopping to stop training when validation loss stops improving.

Results Interpretation

Before: Training accuracy 98%, Validation accuracy 70% (overfitting)

After: Training accuracy 90%, Validation accuracy 86% (better generalization)

Adding dropout and data augmentation helps the model learn general patterns, not just memorize training images. This improves how well the model 'sees' new images.

Bonus Experiment

Try using a smaller learning rate and batch size to see if validation accuracy improves further.

💡 Hint

Lower learning rates can help the model learn more carefully, and smaller batches add noise that can reduce overfitting.