🦾 ML Notebooks

This repository contains Jupyter notebooks covering various deep learning and computer vision topics. Each notebook provides hands-on experience with different techniques and architectures.

---

📚 Notebooks

0️⃣ Intro Labs 🏁

• 🔰 Introduction to Machine Learning concepts:
  • Python
  • NumPy
  • Pandas
  • Data visualization
  • Basic supervised learning
  • Basic semi-supervised learning

1️⃣ Convolutional Neural Networks (CNNs) 🖼️

• 📌 Introduction to Convolutional Neural Networks (CNNs) for image classification.
• 🛠️ Practical implementation using deep learning frameworks.

2️⃣ CNN Architectures 🏗️

• 🔍 Exploration of various CNN architectures.
• 🎛️ Understanding hyperparameter tuning and architectural choices.
• 📦 Utilizing pre-trained models for improved performance.

3️⃣ Encoder-Decoder Transformers 🤖

• 📖 Implementation of the Transformer model for deep learning tasks.
• 🧠 Understanding self-attention mechanisms.

4️⃣ Object Detection 🎯

• 🔍 Training and evaluating object detection models.
• 🏷️ Understanding bounding boxes and class predictions.
• 🛠️ Recommended external resources: Ultralytics for efficient detection models.

5️⃣ Semantic Segmentation 🏞️

• 🎨 Assigning class labels to each pixel in an image.
• 🚗 Applications in medical imaging, autonomous driving, and more.

6️⃣ Semi-Supervised Learning with FixMatch 🏆

• 🏗️ Exploring FixMatch, a method for semi-supervised learning.
• 🖼️ Training on a fraction of CIFAR-10 images.

7️⃣ Self-Supervised Learning with SimCLR 🧩

• 🔄 Implementing SimCLR for contrastive learning.
• 📸 Pretraining on CIFAR-10 and fine-tuning on SVHN.

---

⭐ Getting Started

To run these notebooks locally, install the necessary dependencies:

pip install torch torchvision timm jupyter matplotlib seaborn tqdm

Alternatively, use Google Colab links provided in each section to run them directly in the cloud. ☁️

---

📄 Research Papers

Some of the research papers referenced in the notebooks are available in the pdfs folder. These papers provide theoretical background and additional insights into the concepts explored in the notebooks. These papers complement the practical implementations in this repository and provide additional theoretical insights into the methods used.

📚 Available Papers

• FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence

  • 📄 Found in pdfs/Semi-Supervised Learning FixMatch.pdf

• SimCLR: A Simple Framework for Contrastive Learning of Visual Representations

  • 📄 Found in pdfs/CLR.pdf

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift

  • 📄 Found in pdfs/BatchNormalization.pdf

• MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications

  • 📄 Found in pdfs/MobileNet.pdf

• MobileNetV3: Searching for MobileNetV3

  • 📄 Found in pdfs/MobileNetV3.pdf

• Deep Residual Learning for Image Recognition (ResNet)

  • 📄 Found in pdfs/ResNet.pdf

• Identifying Mislabeled Data using the Area Under the Margin Ranking

  • 📄 Found in pdfs/Mislabeled Data AUC.pdf

• Mixup: Beyond Empirical Risk Minimization

  • 📄 Found in pdfs/RiskMinimization.pdf

• MarginMatch: Improving Semi-Supervised Learning with Pseudo-Margins

  • 📄 Found in pdfs/Semi-Supervised Learning Pseudo-Margins.pdf

• SqueezeNet: AlexNet-Level Accuracy with 50x Fewer Parameters

  • 📄 Found in pdfs/SqueezeNet.pdf

• SSD: Single Shot MultiBox Detector for Object Detection

  • 📄 Found in pdfs/SSD.pdf