Running the Face Scripts with Anaconda (Tutorial)

This tutorial explains how to run two Python scripts that:

1. Align faces in a set of images
2. Draw facial landmark points on those faces

No prior experience with computer vision is required.

---

1. Install Anaconda

1. Go to: https://www.anaconda.com/download
2. Download Anaconda for your operating system (Windows / macOS / Linux)
3. Install it using the default options

After installation, you should have:

• Anaconda Navigator
• Anaconda Prompt (Windows) or Terminal (macOS/Linux)

---

2. Create a New Conda Environment

This keeps everything clean and avoids conflicts.

Open Anaconda Prompt / Terminal, then type:

conda create -n enviroment_name python=3.9

Press Y when asked.

Activate the environment:

conda activate enviroment_name

You should now see (enviroment_name) at the beginning of the line.

---

3. Install Required Libraries

Run these commands one by one:

conda install -c conda-forge dlib
conda install opencv numpy
pip install imutils

⚠️ dlib may take a few minutes to install — this is normal.

---

4. Download the Face Landmark Model

These scripts need a pre-trained face model.

1. Download this file: http://dlib.net/files/shape_predictor_68_face_landmarks.dat.bz2
2. Unzip it
3. Place the file shape_predictor_68_face_landmarks.dat in the same folder as the Python scripts

---

5. Organise the Project Folder

Your folder should look like this:

face-project/
├── face-aligner.py
├── face-points.py
├── face-points-only.py
├── shape_predictor_68_face_landmarks.dat
├── input/
│   ├── image1.jpg
│   ├── image2.jpg
│   └── ...

• Put portrait photos (JPG or PNG) inside the input folder
• Images should contain one face per image

---

6. Run the scripts

In the terminal, navigate to your project folder:

cd path/to/face-project

Then run:

python face-aligner.py

This will:

• Detect faces
• Rotate and center them
• Save the results into a new folder called:

faces/

---

7. Run the Face Points Script

Now run:

python face-points.py

This will:

• Read the aligned faces
• Detect facial landmarks
• Draw colored points on the face
• Save the results into:

faces-points/

face-points-only.py will return only the colored points without the faces.

---

8. Final Output

After everything runs, you should have:

faces/          → normalized portraits  
faces-points/   → portraits with facial landmarks

These images show how a face is interpreted not as an identity, but as a set of measurable points.