Computer Vision (CV) is a field of artificial intelligence that enables machines to interpret and understand visual information from the world — images, videos, and real-time camera feeds.

graph LR
    A["📷 Image/Video Input"] --> B["🔍 Preprocessing"]
    B --> C["🧠 Deep Learning Model"]
    C --> D["📊 Prediction / Classification"]
    D --> E["🎯 Action / Decision"]

    style A fill:#4CAF50,stroke:#333,color:#fff
    style B fill:#2196F3,stroke:#333,color:#fff
    style C fill:#FF9800,stroke:#333,color:#fff
    style D fill:#9C27B0,stroke:#333,color:#fff
    style E fill:#F44336,stroke:#333,color:#fff

computerV/
├── 📄 README.md                  ← You are here
├── 📄 requirements.txt           ← All Python dependencies (libraries needed)
├── 📄 setup.ps1                  ← One-click setup script (Windows PowerShell)
├── 📄 setup.sh                   ← One-click setup script (macOS/Linux Terminal)
├── 📄 .gitignore                 ← Files Git should ignore
│
├── 📂 docs/                      ← Detailed concept guides & tutorials
│   ├── 01_introduction_to_computer_vision.md
│   ├── 02_convolutional_neural_networks.md
│   ├── 03_recurrent_neural_networks.md
│   ├── 04_long_short_term_memory.md
│   ├── 05_cnn_rnn_lstm_combined.md
│   ├── 06_azure_deployment_guide.md
│   ├── 07_linkedin_publishing_guide.md
│   └── 08_run_setup_lean.md
│
├── 📂 src/                       ← All Python source code
│   ├── 📂 01_cnn/
│   │   └── cnn_image_classifier.py    ← CNN model training script
│   ├── 📂 02_rnn/
│   │   └── rnn_sequence_model.py      ← RNN model training script
│   ├── 📂 03_lstm/
│   │   └── lstm_model.py              ← LSTM model training script
│   ├── 📂 04_combined/
│   │   └── cnn_rnn_lstm_combined.py   ← Combined architecture script
│   └── 📂 utils/
│       ├── data_loader.py             ← Dataset loading utilities
│       └── visualization.py           ← Plotting and visualization helpers
│
├── 📂 data/                      ← Datasets (auto-downloaded when you run scripts)
├── 📂 models/                    ← Saved model weights (.pth files)
└── 📂 outputs/                   ← Training plots & metrics (generated during training)

This section provides complete step-by-step instructions for both Windows and macOS/Linux users.

Before you begin, make sure you have these installed:

💡 Tip for Windows users: When installing Python, make sure to check ✅ "Add Python to PATH" during installation!

Open your terminal (or PowerShell on Windows) and run:

git clone https://github.com/EricKart/computerV.git
cd computerV

What this does:

• git clone downloads all the project files from GitHub to your computer
• cd computerV moves you into the project folder

The setup script automatically creates a virtual environment and installs all required packages.

.\setup.ps1

⚠️ If you get an error about "execution policy", run this first:

Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass

Then try .\setup.ps1 again.

chmod +x setup.sh
./setup.sh

What the setup script does:

1. ✅ Creates a venv/ folder (virtual environment) to isolate project dependencies
2. ✅ Upgrades pip (Python's package installer) to the latest version
3. ✅ Installs all required libraries from requirements.txt (PyTorch, NumPy, etc.)
4. ✅ Creates necessary folders: data/, models/, outputs/, logs/

Before running any Python code, you must activate the virtual environment:

.\venv\Scripts\Activate.ps1

You'll see (venv) appear at the beginning of your command line, like this:

(venv) PS C:\Users\YourName\computerV>

source venv/bin/activate

You'll see (venv) appear at the beginning of your command line, like this:

(venv) user@computer:~/computerV$

💡 Remember: You need to activate the virtual environment every time you open a new terminal window to work on this project!

Run these commands to make sure everything is set up correctly:

python --version

Expected output: Python 3.10.x or higher

python -c "import torch, torchvision, matplotlib, onnx; print('✅ Environment OK - All packages installed!')"

Expected output: ✅ Environment OK - All packages installed!

Now you can run any of the training scripts. Start with the CNN module:

python src/01_cnn/cnn_image_classifier.py

Understanding why we use virtual environments is important for every Python developer.

Imagine you have two Python projects:

• Project A needs numpy version 1.20
• Project B needs numpy version 1.24

If you install both globally, they will conflict! Only one version can exist at a time.

A virtual environment is like a separate, isolated Python installation for each project.

Your Computer
├── Global Python (system-wide)
│
├── Project A/
│   └── venv/          ← Has its own numpy 1.20
│       └── numpy 1.20
│
└── Project B/
    └── venv/          ← Has its own numpy 1.24
        └── numpy 1.24

1. setup.sh / setup.ps1 creates the venv/ folder
2. requirements.txt lists all packages and their versions
3. Activating (source venv/bin/activate or .\venv\Scripts\Activate.ps1) tells your terminal to use this project's Python and packages
4. Deactivating (deactivate) returns you to global Python

💡 Pro Tip: The venv/ folder is listed in .gitignore so it's never uploaded to GitHub. Each person creates their own venv/ using the setup script.

Each script in src/ is a complete, runnable program that:

1. Downloads the CIFAR-10 dataset (10 categories of 32x32 images)
2. Builds a neural network model (CNN, RNN, LSTM, or Combined)
3. Trains the model for several epochs
4. Evaluates accuracy on test data
5. Saves the trained model to models/
6. Generates visualization plots in outputs/

Make sure your virtual environment is activated (you see (venv) in your terminal), then:

# Module 1: CNN (Convolutional Neural Network)
python src/01_cnn/cnn_image_classifier.py

# Module 2: RNN (Recurrent Neural Network)
python src/02_rnn/rnn_sequence_model.py

# Module 3: LSTM (Long Short-Term Memory)
python src/03_lstm/lstm_model.py

# Module 4: Combined CNN + LSTM
python src/04_combined/cnn_rnn_lstm_combined.py

When you run a script, you'll see:

1. Dataset download (first time only) — downloads ~170MB of images
2. Training progress — shows loss and accuracy for each epoch
3. Final results — test accuracy and best model saved

Example output:

============================================================
  MODULE 1: CNN IMAGE CLASSIFIER
  Dataset : CIFAR-10  |  Model : CifarCNN
============================================================

Epoch [01/15]  Train Loss: 1.4532  Acc: 47.23%  │  Test Loss: 1.1234  Acc: 60.12%  │  23.5s
Epoch [02/15]  Train Loss: 1.0123  Acc: 63.45%  │  Test Loss: 0.9876  Acc: 65.78%  │  22.1s
...
🏆 Best Test Accuracy: 78.45%

After training, check these folders:

Follow this order for the best learning experience:

graph TD
    A["📖 1. Introduction to CV"] --> B["🔲 2. CNN – See Patterns"]
    B --> C["🔄 3. RNN – Understand Sequences"]
    C --> D["🧠 4. LSTM – Remember Long-Term"]
    D --> E["🔗 5. CNN+RNN+LSTM Combined"]
    E --> F["☁️ 6. Deploy to Azure"]
    F --> G["📢 7. Publish on LinkedIn"]

    style A fill:#E8F5E9,stroke:#4CAF50,color:#1B5E20
    style B fill:#E3F2FD,stroke:#2196F3,color:#0D47A1
    style C fill:#FFF3E0,stroke:#FF9800,color:#E65100
    style D fill:#F3E5F5,stroke:#9C27B0,color:#4A148C
    style E fill:#FCE4EC,stroke:#E91E63,color:#880E4F
    style F fill:#E0F7FA,stroke:#00BCD4,color:#006064
    style G fill:#FFF9C4,stroke:#FFEB3B,color:#F57F17

Goal: Classify images from the CIFAR-10 dataset into 10 categories.

graph LR
    A["Input Image\n32×32×3"] --> B["Conv Layer 1\n+ ReLU"]
    B --> C["MaxPool"]
    C --> D["Conv Layer 2\n+ ReLU"]
    D --> E["MaxPool"]
    E --> F["Flatten"]
    F --> G["Fully Connected"]
    G --> H["Output\n10 classes"]

    style A fill:#4CAF50,stroke:#333,color:#fff
    style B fill:#2196F3,stroke:#333,color:#fff
    style C fill:#03A9F4,stroke:#333,color:#fff
    style D fill:#2196F3,stroke:#333,color:#fff
    style E fill:#03A9F4,stroke:#333,color:#fff
    style F fill:#FF9800,stroke:#333,color:#fff
    style G fill:#9C27B0,stroke:#333,color:#fff
    style H fill:#F44336,stroke:#333,color:#fff

Key Concepts:

• Convolution — Slides a small filter across the image to detect edges, textures, shapes
• Pooling — Reduces spatial dimensions (downsampling) while retaining important features
• ReLU — Non-linear activation: f(x) = max(0, x)

📂 Code: src/01_cnn/cnn_image_classifier.py
📖 Docs: docs/02_convolutional_neural_networks.md

python src/01_cnn/cnn_image_classifier.py

Goal: Process sequential image features to understand temporal/spatial patterns.

graph LR
    A["x₁"] --> B["RNN Cell"]
    B --> C["h₁"]
    C --> D["RNN Cell"]
    A2["x₂"] --> D
    D --> E["h₂"]
    E --> F["RNN Cell"]
    A3["x₃"] --> F
    F --> G["h₃"]
    G --> H["Output"]

    style B fill:#FF9800,stroke:#333,color:#fff
    style D fill:#FF9800,stroke:#333,color:#fff
    style F fill:#FF9800,stroke:#333,color:#fff
    style H fill:#F44336,stroke:#333,color:#fff

Key Concepts:

• Hidden State — Memory that carries information from previous time steps
• Recurrence — Same weights applied at every time step
• Backpropagation Through Time (BPTT) — Training across sequences

📂 Code: src/02_rnn/rnn_sequence_model.py
📖 Docs: docs/03_recurrent_neural_networks.md

python src/02_rnn/rnn_sequence_model.py

Goal: Solve the vanishing gradient problem and model long-range dependencies.

graph TD
    subgraph LSTM Cell
        A["Forget Gate 🚪"] --> D["Cell State"]
        B["Input Gate 🚪"] --> D
        D --> E["Output Gate 🚪"]
        E --> F["Hidden State"]
    end

    style A fill:#F44336,stroke:#333,color:#fff
    style B fill:#4CAF50,stroke:#333,color:#fff
    style D fill:#2196F3,stroke:#333,color:#fff
    style E fill:#FF9800,stroke:#333,color:#fff
    style F fill:#9C27B0,stroke:#333,color:#fff

Key Concepts:

• Forget Gate — Decides what information to discard from cell state
• Input Gate — Decides what new information to store
• Output Gate — Decides what to output based on cell state
• Cell State — Long-term memory highway

📂 Code: src/03_lstm/lstm_model.py
📖 Docs: docs/04_long_short_term_memory.md

python src/03_lstm/lstm_model.py

Goal: Build a video classification pipeline — CNN extracts spatial features per frame, LSTM captures temporal patterns across frames.

graph TD
    subgraph "Per-Frame Feature Extraction (CNN)"
        A["Frame 1 🖼️"] --> B["CNN"]
        A2["Frame 2 🖼️"] --> B2["CNN"]
        A3["Frame 3 🖼️"] --> B3["CNN"]
        AN["Frame N 🖼️"] --> BN["CNN"]
    end

    B --> C["Feature Vector 1"]
    B2 --> C2["Feature Vector 2"]
    B3 --> C3["Feature Vector 3"]
    BN --> CN["Feature Vector N"]

    subgraph "Temporal Modeling (LSTM)"
        C --> D["LSTM"]
        C2 --> D
        C3 --> D
        CN --> D
        D --> E["Final Hidden State"]
    end

    E --> F["🎯 Video Class Prediction"]

    style B fill:#2196F3,stroke:#333,color:#fff
    style B2 fill:#2196F3,stroke:#333,color:#fff
    style B3 fill:#2196F3,stroke:#333,color:#fff
    style BN fill:#2196F3,stroke:#333,color:#fff
    style D fill:#FF9800,stroke:#333,color:#fff
    style F fill:#F44336,stroke:#333,color:#fff

📂 Code: src/04_combined/cnn_rnn_lstm_combined.py
📖 Docs: docs/05_cnn_rnn_lstm_combined.md

python src/04_combined/cnn_rnn_lstm_combined.py

Take your trained model from local to production in the cloud.

graph LR
    A["Trained Model\n(.pth)"] --> B["Export to ONNX"]
    B --> C["Azure ML\nWorkspace"]
    C --> D["Deploy as\nREST API"]
    D --> E["Consume from\nAny App"]

    style A fill:#4CAF50,stroke:#333,color:#fff
    style B fill:#2196F3,stroke:#333,color:#fff
    style C fill:#0078D4,stroke:#333,color:#fff
    style D fill:#FF9800,stroke:#333,color:#fff
    style E fill:#9C27B0,stroke:#333,color:#fff

📖 Full Guide: docs/06_azure_deployment_guide.md

Showcase your completed project to recruiters and the tech community.

📖 Full Guide: docs/07_linkedin_publishing_guide.md

If something is broken and you want to start over:

# Delete the virtual environment
rm -rf venv          # macOS/Linux
rmdir /s /q venv     # Windows

# Run setup again
./setup.sh           # macOS/Linux
.\setup.ps1          # Windows

# Pull the latest changes (--ff-only prevents merge commits)
git pull --ff-only

# Re-run setup to install any new dependencies
./setup.sh           # macOS/Linux
.\setup.ps1          # Windows

1. Fork the repository
2. Create a feature branch: git checkout -b feature/amazing-addition
3. Commit your changes: git commit -m 'Add amazing addition'
4. Push to the branch: git push origin feature/amazing-addition
5. Open a Pull Request

This project is licensed under the MIT License — see the LICENSE file for details.