Added image tester class, provided testing example

.gitignore

@@ -136,4 +136,6 @@ dmypy.json
 Data*
 **/Data*
 
-Save*
+Save*
+
+test_images

image_recognition/cli.py

@@ -2,7 +2,8 @@
 
 from image_recognition.modules.importer import importer
 from image_recognition.modules.model import model
-from image_recognition.modules.visualization import plotter_evaluator
+from image_recognition.modules.visualization import plotter_evaluator, tester
+import os
 
 def main():  # pragma: no cover
 
@@ -38,6 +39,7 @@ def main():  # pragma: no cover
 
     ## Обработка данных
     # Вносится рандомизация (ротация, зум, перемещение). Также приводится яркость к понятному нейросети формату (вместо 0-255, 0-1).
+    # По-умолчанию (0.2, 0.1, 0.08)
     i.generate_augmentation_layers(0.2, 0.1, 0.08)
 
     ### Применение слоёв обработки данных
@@ -104,3 +106,17 @@ def main():  # pragma: no cover
     ## Матрица запутанности
     # Хорший способ понять, как именно нейросеть ошибается
     pe.plot_confusion_matrix()
+
+
+    # Тестирование избранных случаев
+    t = tester(model_instance)
+
+    images = ["one", "two", "tree", "four", "five", "six", "seven", "eight", "nine", "zero"]
+    # Составной путь. Строится от корня программы
+    for i in range(len(images)):
+
+        absolute_path = os.path.join(os.getcwd(), 'test_images', images[i] + '.jpg')
+        t.read_image(absolute_path, False)
+
+        t.img_predict()
+

image_recognition/modules/importer.py

@@ -71,4 +71,5 @@ def __get_stats(self):
             labels = np.concatenate([labels, np.argmax(y.numpy(), axis=-1)])
 
         self.class_names = set(labels)
-        self.num_classes = len(self.class_names)
+        self.num_classes = len(self.class_names)
+

image_recognition/modules/model.py

@@ -45,7 +45,6 @@ def compile(self):
         # Компиляция модели
         optimizer = RMSprop(learning_rate=0.001, rho=0.9, epsilon=1e-08)
         self.model.compile(optimizer = optimizer , loss = "categorical_crossentropy", metrics=["accuracy"])
-        # self.model.summary()
 
     def train(self, train_ds: tf_data.Dataset, epochs: int, validation_data: tf_data.Dataset):
 
@@ -65,4 +64,5 @@ def init_learning_rate_reduction(self):
 
     def init_save_at_epoch(self):
 
-        self.callbacks.append(keras.callbacks.ModelCheckpoint("save_at_{epoch}.keras"))
+        self.callbacks.append(keras.callbacks.ModelCheckpoint("save_at_{epoch}.keras"))
+

image_recognition/modules/visualization.py

@@ -9,6 +9,7 @@
 from scikitplot.metrics import plot_confusion_matrix 
 from tensorflow import data as tf_data
 
+
 class plotter_evaluator:
 
     model: keras.Model
@@ -62,4 +63,40 @@ def print_report(self):
     def plot_confusion_matrix(self):
 
         plot_confusion_matrix(self.labels, self.pred_labels,cmap= 'YlGnBu')
-        plt.show()
+        plt.show()
+
+
+import cv2
+import tensorflow as tf
+import math
+
+
+class tester:
+
+    img: np.ndarray | None = None
+    model: keras.Model
+    path: str
+
+    def __init__(self, model: keras.Model):
+        self.model = model
+
+    def read_image(self, path: str, show: bool):
+        img = cv2.imread(path, cv2.IMREAD_GRAYSCALE)
+        if(img is None):
+            print("Image not found")
+            return
+
+        if(show): plt.imshow(img, cmap='gray', vmin=0, vmax=255)
+        self.path = path
+        self.img = img
+
+    def img_predict(self):
+        if(self.img is None):
+            print("No image supplied")
+            return
+
+        prediction = self.model.predict(np.expand_dims(self.img/255, 0))
+        probabilities = list(map(lambda x: math.floor(x*1000)/1000, prediction[0]))
+        print(self.path)
+        print(probabilities)
+