Unit 8 Unit 11/1

Unit 9 - CNN Model Activity

This unit explored the use of convolutional neural networks (CNNs) for object recognition and the ethical implications of such technologies. Practical experimentation involved modifying the input image to observe the model's prediction accuracy.

Key Learning Outcomes

  • Legal, Social, and Ethical Issues: Reviewed ethical concerns highlighted in Wall (2019), including privacy, misuse of facial recognition, and biases embedded in training datasets. Discussed the potential for social harm if CNNs are deployed without proper accountability or fairness checks.
  • Applicability and Dataset Challenges: Highlighted the importance of dataset quality and diversity to avoid biased predictions. Observed model accuracy for different input images, noting that some indices (e.g., [1], [5]) produced correct classifications, while others resulted in misclassifications, underscoring dataset limitations.

Key Artefacts

  • Modified Predictions: Experimented with input indices ([1–15]) to assess model consistency and accuracy.
  • Discussion of Bias: Documented the effects of dataset bias on prediction performance and ethical considerations.
  • Code Implementation: Reviewed and analyzed CNN architecture and performance metrics.

Self-Reflection

  • Strengths: Developed a deeper understanding of CNN architecture and its application to object recognition tasks.
  • Improvements: Gained awareness of ethical considerations in deploying CNN models, with a focus on reducing biases and ensuring responsible use.

Code Showcase

						
# Example CNN implementation using TensorFlow
import tensorflow as tf
from tensorflow.keras import layers, models

# Define the CNN model
model = models.Sequential([
    layers.Conv2D(32, (3, 3), activation='relu', input_shape=(32, 32, 3)),
    layers.MaxPooling2D((2, 2)),
    layers.Conv2D(64, (3, 3), activation='relu'),
    layers.MaxPooling2D((2, 2)),
    layers.Flatten(),
    layers.Dense(64, activation='relu'),
    layers.Dense(10, activation='softmax')
])

# Compile and train the model
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])
model.fit(x_train, y_train, epochs=10, validation_data=(x_test, y_test))