implement ML model conversion, deploy models, and fix card detection noise

.gitignore

@@ -1,3 +1,7 @@
+/venv
+/dataset
+*.h5
+
 # Dependencies
 node_modules/
 .pnp/

ML_SETUP_GUIDE.md

@@ -12,7 +12,7 @@ The system uses a **Hybrid Pipeline** to balance performance on mobile devices w
 2.  **Classification**: Crops each detected card and passes it through two specialized TensorFlow.js models to determine the **Suit** and the **Value**.
 
 ## 2. Data Collection & Labeling
-Because Jass cards have unique iconography, we use a custom dataset created from internet samples.
+Because Jass cards have unique iconography, we use a custom dataset.
 
 ### Labeling Strategy: Folder-Based Annotation
 Instead of using bounding-box tools, we use the directory structure as labels.
@@ -29,7 +29,7 @@ dataset/
     ├── 6/
     ├── 7/
     ...
-    └── 13/
+    └── 14/
 ```
 
 ### Process:

convert_model.sh

@@ -0,0 +1,32 @@
+#!/bin/bash
+
+# Check if tensorflowjs is installed
+if ! command -v tensorflowjs_converter &> /dev/null
+then
+    echo "tensorflowjs_converter could not be found. Please install it using: pip install tensorflowjs"
+    exit 1
+fi
+
+MODEL_TYPE=$1
+INPUT_MODEL=$2
+OUTPUT_DIR=$3
+
+if [ -z "$MODEL_TYPE" ] || [ -z "$INPUT_MODEL" ] || [ -z "$OUTPUT_DIR" ]; then
+    echo "Usage: ./convert_model.sh [suit|value] [path_to_h5_model] [output_directory]"
+    exit 1
+fi
+
+echo "Converting $MODEL_TYPE model from $INPUT_MODEL to $OUTPUT_DIR..."
+
+# Create output directory if it doesn't exist
+mkdir -p "$OUTPUT_DIR"
+
+# Perform conversion
+tensorflowjs_converter --input_format=keras "$INPUT_MODEL" "$OUTPUT_DIR"
+
+if [ $? -eq 0 ]; then
+    echo "Conversion successful. Model saved to $OUTPUT_DIR"
+else
+    echo "Conversion failed."
+    exit 1
+fi

src/App.jsx

@@ -18,15 +18,6 @@ const App = () => {
   );
 };
 
-
-const App = () => {
-  return (
-    <GameStateProvider>
-      <AppContent />
-    </GameStateProvider>
-  );
-};
-
 const AppContent = () => {
   const { gameState } = useGameStateContext();
 

src/components/Detection/Detection.tsx

@@ -98,9 +98,8 @@ const Detection: React.FC<DetectionProps> = ({ videoRef, canvasRef, onCardsDetec
   // Enhanced card region detection specialized for Jass cards
     const findCardRegions = (imageData: ImageData, width: number, height: number): {x: number, y: number, width: number, height: number}[] => {
       const regions = [];
-    const step = 12; // Adjusted step for better detection
+      const step = 24; // Increased step to reduce noise and overlapping detections
       
-    // Look for card-colored regions: typically white/gray cards with identifiable suit symbols
       for (let y = 0; y < height; y += step) {
         for (let x = 0; x < width; x += step) {
           const i = (y * width + x) * 4;
@@ -109,17 +108,11 @@ const Detection: React.FC<DetectionProps> = ({ videoRef, canvasRef, onCardsDetec
           const b = imageData.data[i + 2];
           const brightness = (r + g + b) / 3;
           
-        // Card background brightness range (light cards)
-        if (brightness > 180 && brightness < 250) {
+          if (brightness > 200 && brightness < 255) {
             const region = getCardRegionWithShapeAnalysis(imageData, width, height, x, y, step);
-          if (region && region.width > 30 && region.height > 40) {
-            // Check that this is a proper rectangular card area by testing aspect ratio
-            const widthDiff = region.width;
-            const heightDiff = region.height;
-            const aspectRatio = widthDiff / heightDiff;
-            
-            // Typical playing card aspect ratio (roughly 0.7 to 1.3)
-            if (aspectRatio > 0.7 && aspectRatio < 1.3) {
+            if (region && region.width > 60 && region.height > 80) {
+              const aspectRatio = region.width / region.height;
+              if (aspectRatio > 0.6 && aspectRatio < 1.4) {
                 regions.push(region);
               }
             }
@@ -127,7 +120,20 @@ const Detection: React.FC<DetectionProps> = ({ videoRef, canvasRef, onCardsDetec
         }
       }
       
-    return regions;
+      // Remove overlapping regions to avoid "dozens of nonsense cards"
+      const uniqueRegions = [];
+      regions.sort((a, b) => (a.width * a.height) - (b.width * b.height));
+      
+      for (const region of regions) {
+        const isOverlapping = uniqueRegions.some(u => 
+          Math.abs(u.x - region.x) < 30 && Math.abs(u.y - region.y) < 30
+        );
+        if (!isOverlapping) {
+          uniqueRegions.push(region);
+        }
+      }
+      
+      return uniqueRegions;
     };
 
   // Improved card region extraction with shape analysis for more accurate detection

src/components/Results/ResultsScreen.tsx

@@ -80,7 +80,12 @@ const ResultsScreen: React.FC = () => {
             <div className="cards-grid">
               {gameState.detectedCards.map(card => (
                 <div key={card.id} className="card-preview">
-                <div className="card-suit">♠</div>
+                  <div className="card-suit">
+                    {card.suit === 'Schellen' && '🔔'}
+                    {card.suit === 'Schilten' && '🛡️'}
+                    {card.suit === 'Eicheln' && '🌰'}
+                    {card.suit === 'Rosen' && '🌹'}
+                  </div>
                   <div className="card-value">{card.value || '?'}pts</div>
                 </div>
               ))}

src/services/CardModelService.ts

@@ -13,7 +13,7 @@ class CardModelService {
   private readonly VALUE_MODEL_PATH = '/models/value_model/model.json';
 
   private readonly SUIT_LABELS = ['Schellen', 'Schilten', 'Eicheln', 'Rosen'];
-  private readonly VALUE_LABELS = ['6', '7', '8', '9', '10', '11', '12', '13'];
+  private readonly VALUE_LABELS = ['6', '7', '8', '9', '10', '11', '12', '13', '14'];
 
   async init(): Promise<void> {
     try {

train_model.py

@@ -0,0 +1,82 @@
+import os
+import tensorflow as tf
+from tensorflow.keras import layers, models, optimizers
+from tensorflow.keras.preprocessing.image import ImageDataGenerator
+
+def create_model(num_classes):
+    # Use MobileNetV2 as the base model
+    base_model = tf.keras.applications.MobileNetV2(
+        input_shape=(64, 64, 3),
+        include_top=False,
+        weights='imagenet'
+    )
+    base_model.trainable = False  # Freeze base model for transfer learning
+
+    model = models.Sequential([
+        base_model,
+        layers.GlobalAveragePooling2D(),
+        layers.Dropout(0.2),
+        layers.Dense(num_classes, activation='softmax')
+    ])
+
+    model.compile(
+        optimizer=optimizers.Adam(),
+        loss='categorical_crossentropy',
+        metrics=['accuracy']
+    )
+    return model
+
+def train(model_type, dataset_root, output_path):
+    # Set parameters based on model type
+    if model_type == 'suit':
+        num_classes = 4
+    elif model_type == 'value':
+        num_classes = 9
+    else:
+        raise ValueError("model_type must be 'suit' or 'value'")
+
+    # Data augmentation as per ML_SETUP_GUIDE.md
+    datagen = ImageDataGenerator(
+        rescale=1./255,
+        rotation_range=15,
+        brightness_range=[0.8, 1.2],
+        validation_split=0.2
+    )
+
+    train_generator = datagen.flow_from_directory(
+        os.path.join(dataset_root, f'{model_type}_model'),
+        target_size=(64, 64),
+        batch_size=32,
+        class_mode='categorical',
+        subset='training'
+    )
+
+    validation_generator = datagen.flow_from_directory(
+        os.path.join(dataset_root, f'{model_type}_model'),
+        target_size=(64, 64),
+        batch_size=32,
+        class_mode='categorical',
+        subset='validation'
+    )
+
+    model = create_model(num_classes)
+    
+    print(f"Training {model_type} model...")
+    model.fit(
+        train_generator,
+        epochs=20,
+        validation_data=validation_generator
+    )
+
+    model.save(output_path)
+    print(f"Model saved to {output_path}")
+
+if __name__ == "__main__":
+    import argparse
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--type', choices=['suit', 'value'], required=True)
+    parser.add_argument('--dataset', default='dataset')
+    parser.add_argument('--output', required=True)
+    args = parser.parse_args()
+
+    train(args.type, args.dataset, args.output)