Deep Learning Projects Portfolio 🧠

A collection of end-to-end Deep Learning architectures, ranging from custom Convolutional Neural Networks built from scratch to Transfer Learning pipelines using state-of-the-art pre-trained models.

🚀 Key Projects

1. Transfer Learning: Cat vs. Dog Classifier (VGG16)

• File: Transfer_Learning_VGG16_Cats_Dogs_assignment...
• Description: Leveraged the pre-trained VGG16 model (ImageNet weights) as a feature extractor. Built and trained a custom fully connected classification head, achieving 100% confidence on test images using real-time data augmentation.

2. Custom CNN: Cat vs. Dog Image Classification

• File: CNN_Cats_vs_Dogs.ipynb
• Description: Engineered a custom Convolutional Neural Network from scratch. Implemented image data generators, dropout for regularization, and achieved solid validation accuracy on a complex visual dataset.

3. Fashion MNIST: Multi-Class Image Classification

• File: Fashion_mnist_ANN_assignment_6_DL_Intellipa...
• Description: Built a fully connected Artificial Neural Network (ANN) using Keras to classify Zalando's clothing dataset into 10 distinct apparel categories. Implemented data normalization and TensorBoard for visualization.

4. MNIST Digits & Neural Network Foundations

• Files: Mnist_Digits Fully Connected Neural Network.ipynb, manual_small_neural_network.ipynb, DLbasics.ipynb
• Description: Core foundational projects exploring the math and mechanics of neural networks, including manual network construction and baseline digit recognition.

🛠️ Tech Stack

• Frameworks: TensorFlow, Keras
• Libraries: NumPy, Pandas, Matplotlib
• Techniques: CNNs, Transfer Learning (VGG16), Data Augmentation, TensorBoard, Hyperparameter Tuning