WIP

longinus/longinus/coeff_encoders/encoder.py

@@ -1,7 +0,0 @@
-class Encoder:
-    def __init__(self, encoder):
-        self.encoder = encoder
-    
-    def train_model(self, dataset):
-
-        pass

longinus/longinus/coeff_encoders/neural_network.py

@@ -1,3 +1,32 @@
+from keras.models import Sequential, load_model
+from keras.layers import Conv2D, MaxPooling2D, Flatten, Dense
+import pandas as pd
+
+
 class NeuralNetworkEncoder:
-    def __init__(self):
-        pass
+    def __init__(self, internal_activation_function: str, external_activation_function: str, optimizer: str, loss_function: str, image_dimension_x: int, image_dimension_y: int):
+        self.model = Sequential()
+        self.model.add(Conv2D(32, (3, 3), activation=internal_activation_function, input_shape=(image_dimension_x, image_dimension_y, 1)))
+        self.model.add(MaxPooling2D((2, 2)))
+        self.model.add(Conv2D(64, (3, 3), activation=internal_activation_function))
+        self.model.add(MaxPooling2D((2, 2)))
+        self.model.add(Flatten())
+        self.model.add(Dense(64, activation=internal_activation_function))
+        self.model.add(Dense((8, 8), activation=external_activation_function))
+        self.model.compile(optimizer=optimizer, loss=loss_function)
+    
+    def __init__(self, pretrained_weights_path: str):
+        self.model = load_model(pretrained_weights_path)
+
+    def train(self, train_imageset: pd.DataFrame, train_quantization_dataset: pd.DataFrame, validation_dataset: pd.DataFrame, validation_quantization_dataset: pd.DataFrame, epochs: int, batch_size: int):
+        self.model.fit(train_imageset, train_quantization_dataset, validation_data=(validation_dataset, validation_quantization_dataset), epochs=epochs, batch_size=batch_size)
+
+    def export_weights(self, output_path: str):
+        self.model.save(output_path)
+
+    def evaluate(self, test_imageset: pd.DataFrame, test_quantization_dataset: pd.DataFrame):
+        return self.model.evaluate(test_imageset, test_quantization_dataset)
+
+    def predict_quantization_table(self, imageset: pd.DataFrame):
+        return self.model.predict(imageset)
+    

longinus/longinus/coeff_encoders/random_forest.py

@@ -1,19 +0,0 @@
-import numpy as np
-from sklearn.ensemble import RandomForestRegressor
-from sklearn.model_selection import cross_val_score
-from sklearn.model_selection import train_test_split
-
-class RandomForestEncoder:
-    def __init__(self):
-        self.engine = RandomForestRegressor(n_estimators=100, max_depth=10)
-        self.dataset_training_X = None
-        self.dataset_training_y = None
-        self.dataset_testing_X = None
-        self.dataset_testing_y = None
-
-    def train_model(self, dataset):
-        self.dataset_training_X, self.dataset_training_y, self.dataset_testing_X, self.dataset_testing_y = train_test_split(dataset, test_size=0.2)
-        self.engine.fit(self.dataset_training_X, self.dataset_training_y)
-    
-    def evaluate_model(self):
-        return cross_val_score(self.engine, self.dataset_testing_X, self.dataset_testing_y)

longinus/longinus/utils/dataset.py

@@ -32,6 +32,16 @@ class Dataset():
                 self.dataset_filenames.append(new_filename)
         self.dataset_size = counter
     
+    def load_prepared_dataset(self):
+        for filename in os.listdir(self.dataset_path):
+            if os.path.isfile(file):
+                counter += 1
+            else:
+                continue
+            file = os.path.join(self.dataset_path, filename)
+            self.dataset_filenames.append(file)
+            
+
     def __len__(self):
         return self.dataset_size
     

longinus/pyproject.toml

@@ -17,6 +17,8 @@ pillow = "^10.2.0"
 scipy = "^1.12.0"
 scikit-learn = "^1.4.1.post1"
 opencv-python = "^4.9.0.80"
+keras = "^3.3.3"
+keras-models = "^0.0.7"
 
 
 [build-system]