A comprehensive collection of hands-on notebooks designed to guide beginners through fundamental deep learning concepts and practical implementations.
Overview • Learning Path • Quick Start • Models • Prerequisites
Reason behind Deep Learning Beginner - the reason I have developed this project is that, I used to have a lot of headache how I should deep-dive into the deep learning field. I was not even sure that, what was the very first step for it. So, now, I saved my first approach on entering deep learning.
Why Numpy - First of all ===> The fundamental building block of the deep learning is numpy, we can say how neural networks are constructed and mathematically processed are directly related with numpy theories.
What are Models & Data - Secondly ===> You must choose the Tensorflow ( OR ) Pytorch. If your choice is Tensorflow, you are more aligned with this repo. I have built image-classification on labels ( technically, what is inside the image) by using the built-in datasets of datasets provided via TensorFlow’s Keras API.
- NumPy Fundamentals: Master array operations, broadcasting, and mathematical functions essential for deep learning
- Neural Network Basics: Understand the building blocks of neural networks through practical examples
- Image Classification: Build very basic deep learning models by using Tensorflow built-in image datasets
- TensorFlow/Keras: Get comfortable with the leading deep learning framework
- Model Training Pipeline: Learn data preprocessing, model building, training, and evaluation
Apply your knowledge by building real image classification models.
| Model | Dataset | Task | Data-Complexity |
|---|---|---|---|
| MNIST | Handwritten Digits | 10-class classification | ⭐ |
| Fashion-MNIST | Clothing Items | 10-class classification | ⭐⭐ |
| CIFAR-10 | Natural Images | 10-class classification | ⭐⭐⭐ |
1️⃣ Clone or Download this Repository
git clone https://github.com/thonedra-dev/deeplearning_beginner.git2️⃣ Upload to Google Drive (Optional)
- Upload the repository folder to your Google Drive for easy access
3️⃣ Open in Google Colab
- Navigate to Google Colab
- Select
File→Upload notebookorOpen notebook→GitHub - Upload individual
.pyor.ipynbfiles
File: mnist_model.py
The classic starting point for image classification. Learn to build a neural network that recognizes handwritten digits (0-9).
What You'll Build:
- Simple feedforward neural network
- Convolutional Neural Network (CNN)
- Model evaluation and visualization
Dataset: 60,000 training images + 10,000 test images (28×28 grayscale)
File: fashion_mnist.py
Step up the challenge with more complex grayscale images of clothing items.
What You'll Build:
- Enhanced CNN architecture
- Data augmentation techniques
- Performance optimization strategies
Dataset: 10 classes (T-shirt, Trouser, Pullover, Dress, Coat, Sandal, Shirt, Sneaker, Bag, Ankle boot)
File: ciphar10_model.py
Master color image classification with real-world complexity.
What You'll Build:
- Deep CNN with multiple layers
- Batch normalization and dropout
- Advanced training techniques
Dataset: 60,000 color images (32×32 RGB) across 10 classes (airplane, car, bird, cat, deer, dog, frog, horse, ship, truck)
- Basic Python programming
- Understanding of functions, loops, and data structures
- Python 3.9 or higher
- Google Colab account (free) OR
- Jupyter Notebook installed locally
numpy>=1.21.0
tensorflow>=2.10.0
matplotlib>=3.5.0💡 Pro Tip: Google Colab comes with all these libraries pre-installed!
deeplearning_beginner/
├── numpy_practice/
│ ├── day_one_one.py
│ ├── day_one_two.py
│ ├── day_one_three.py
│ ├── day_one_four.py
│ ├── day_one_five.py
│ ├── day_two_one.py
│ ├── day_two_two.py
│ ├── day_two_three.py
│ ├── day_two_four.py
│ ├── day_three_one.py
│ ├── day_three_two.py
│ └── day_three_three.py
│
├── model_notebooks/
│ ├── mnist_model.ipynb
│ ├── fashion_mnist.ipynb
│ └── cifar10_model.ipynb
│
├── requirements.txt
└── README.md
- 📧 Email me with questions: thonedra.dev@gmail.com
- 🐛 Report issues in the Issues section
- 💬 Start a discussion in Discussions
Use this checklist to track your learning journey:
- Completed all Day 1 notebooks (NumPy basics)
- Completed all Day 2 notebooks (Array operations)
- Completed all Day 3 notebooks (Advanced concepts)
- Built and trained MNIST model
- Built and trained Fashion-MNIST model
- Built and trained CIFAR-10 model
- Achieved >95% accuracy on MNIST
- Achieved >85% accuracy on Fashion-MNIST
- Achieved >70% accuracy on CIFAR-10
