gists/train_model_to_recognize_patterns.md

# Model recognition

## Training source code:
```python                                                                                                                                              
import cv2                                                                 
import numpy as np                                                         
from tensorflow.keras.models import Sequential                                                                                                         
from tensorflow.keras.layers import Dense, Dropout                                                                                                     
                                     
# Load dataset of images with desired output (facial expressions)                                                                                      
data_dir = 'path/to/dataset'                                                                                                                           
image_paths = [f for f in os.listdir(data_dir) if f.endswith('.jpg')]                                                                                  
                                                                           
# Create a list to hold our image data and labels                                                                                                      
X_train = []                                                               
y_train = []                                                                                                                                           
                                                                           
for i, img_path in enumerate(image_paths):                                                                                                             
    # Load image from disk using OpenCV                                    
    img = cv2.imread(os.path.join(data_dir, img_path))                                                                                                 
                                                                                                                                                       
    # Resize image to 224x224 pixels (standard for VGG16)                                                                                              
    img = cv2.resize(img, (224, 224))                                                                                                                  
                                                                                                                                                       
    # Convert image to grayscale and normalize pixel values                                                                                            
    img_gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)                                                                                                   
    img_norm = img_gray / 255.0                                                                                                                        
                                     
    # Add image data to list                                                                                                                           
    X_train.append(np.array(img_norm))                                     
                                                                                                                                                       
    # Assign label for current facial expression (e.g., happy, sad, angry)                                                                             
    if i % 3 == 0:  # Label for 'happy' expression                                                                                                     
        y_train.append(1)                                                  
    elif i % 3 == 1:  # Label for 'sad' expression                                                                                                     
        y_train.append(2)                                                  
    else:  # Label for 'angry' expression                                                                                                              
        y_train.append(3)                        
    else:  # Label for 'angry' expression                                                                                   
        y_train.append(3)                                                                                                                              
                                                                                                                                                       
# Create a TensorFlow model using the VGG16 architecture as base                                                                                       
model = Sequential()                                                       
model.add(Dense(64, activation='relu', input_shape=(224*224,)))                                                                                        
model.add(Dropout(0.2))                                                    
model.add(Dense(32, activation='relu'))                                    
model.add(Dropout(0.2))                                                                                                                                
model.add(Dense(3))  # Output layer for 3 facial expressions                                                                                           
                                     
# Compile the model with suitable loss function and optimizer                                                                                          
model.compile(loss='sparse_categorical_crossentropy',                                                                                                  
              optimizer='adam',                                                                                                                        
              metrics=['accuracy'])                                        
                                                                                                                                                       
# Train the model using our dataset                                        
model.fit(np.array(X_train), np.array(y_train), epochs=10, batch_size=32)                                                                              
                                                                           
print("Model trained!")                              
```