From c80549d5d3eb8e579f54502655ac42d543bd53ad Mon Sep 17 00:00:00 2001
From: Matthias Bisping <matthias.bisping@iqser.com>
Date: Fri, 1 Apr 2022 21:32:18 +0200
Subject: [PATCH] refactoring: model wrapper to base class and derived class
 for efficient net

---
 .../redai_adapter/efficient_net_wrapper.py    | 50 +++++++++++++++++++
 image_prediction/redai_adapter/mlflow.py      |  1 +
 .../redai_adapter/model_wrapper.py            | 42 ++++------------
 3 files changed, 60 insertions(+), 33 deletions(-)
 create mode 100644 image_prediction/redai_adapter/efficient_net_wrapper.py

diff --git a/image_prediction/redai_adapter/efficient_net_wrapper.py b/image_prediction/redai_adapter/efficient_net_wrapper.py
new file mode 100644
index 0000000..f43a1b7
--- /dev/null
+++ b/image_prediction/redai_adapter/efficient_net_wrapper.py
@@ -0,0 +1,50 @@
+import os
+
+from image_prediction.redai_adapter.model_wrapper import ModelWrapper
+
+os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"
+
+import tensorflow as tf
+
+
+class EfficientNetWrapper(ModelWrapper):
+
+    def __init__(self, classes, base_weights_path=None, weights_path=None):
+        self.__input_shape = (224, 224, 3)
+        super().__init__(classes=classes, base_weights_path=base_weights_path, weights_path=weights_path)
+
+    @property
+    def input_shape(self):
+        return self.__input_shape
+
+    def _ModelWrapper__preprocess_tensor(self, tensor):
+        return tf.keras.applications.efficientnet.preprocess_input(tensor)
+
+    def _ModelWrapper__build(self, base_weights=None) -> tf.keras.models.Model:
+        input_img = tf.keras.layers.Input(shape=self.input_shape)
+
+        pretrained = tf.keras.applications.efficientnet.EfficientNetB0(
+            include_top=False, input_tensor=tf.keras.layers.Input(shape=self.input_shape), weights=base_weights
+        )
+
+        pretrained.trainable = False
+
+        for layer in pretrained.layers:
+            layer.trainable = False
+
+        pretrained = pretrained(input_img)
+
+        finetuned = tf.keras.layers.Flatten()(pretrained)
+        finetuned = tf.keras.layers.Dense(512, activation="relu")(finetuned)
+        finetuned = tf.keras.layers.Dropout(0.2)(finetuned)
+        finetuned = tf.keras.layers.Dense(128, activation="relu")(finetuned)
+        finetuned = tf.keras.layers.Dropout(0.2)(finetuned)
+        finetuned = tf.keras.layers.Dense(32, activation="relu")(finetuned)
+        finetuned = tf.keras.layers.Dropout(0.2)(finetuned)
+        finetuned = tf.keras.layers.Dense(len(self.classes), activation="softmax")(finetuned)
+
+        model = tf.keras.models.Model(inputs=input_img, outputs=finetuned)
+
+        model.compile()
+
+        return model
diff --git a/image_prediction/redai_adapter/mlflow.py b/image_prediction/redai_adapter/mlflow.py
index 7930ae3..55ceab3 100644
--- a/image_prediction/redai_adapter/mlflow.py
+++ b/image_prediction/redai_adapter/mlflow.py
@@ -31,6 +31,7 @@ class MlflowModelReader:
 
     @lru_cache(maxsize=None)
     def __get_run(self, run_id):
+
         return mlflow.get_run(run_id)
 
     def __get_classes(self, run_id, prefix="tt"):
diff --git a/image_prediction/redai_adapter/model_wrapper.py b/image_prediction/redai_adapter/model_wrapper.py
index 5bb9a70..7a50f53 100644
--- a/image_prediction/redai_adapter/model_wrapper.py
+++ b/image_prediction/redai_adapter/model_wrapper.py
@@ -1,3 +1,4 @@
+import abc
 import os
 
 os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"
@@ -7,25 +8,25 @@ import numpy as np
 import tensorflow as tf
 
 
-class EfficientNetWrapper:
+class ModelWrapper(abc.ABC):
 
     def __init__(self, classes, base_weights_path=None, weights_path=None):
         self.__classes = classes
-        self.__input_shape = (224, 224, 3)
         self.model = self.__build(base_weights_path)
         self.model.load_weights(weights_path)
 
     @property
+    @abc.abstractmethod
     def input_shape(self):
-        return self.__input_shape
+        raise NotImplementedError
 
     @property
     def classes(self):
         return self.__classes
 
-    @staticmethod
-    def __preprocess_tensor(tensor):
-        return tf.keras.applications.efficientnet.preprocess_input(tensor)
+    @abc.abstractmethod
+    def __preprocess_tensor(self, tensor):
+        raise NotImplementedError
 
     @staticmethod
     def __images_to_tensor(images):
@@ -41,31 +42,6 @@ class EfficientNetWrapper:
 
         return tensor
 
+    @abc.abstractmethod
     def __build(self, base_weights=None) -> tf.keras.models.Model:
-        input_img = tf.keras.layers.Input(shape=self.input_shape)
-
-        pretrained = tf.keras.applications.efficientnet.EfficientNetB0(
-            include_top=False, input_tensor=tf.keras.layers.Input(shape=self.input_shape), weights=base_weights
-        )
-
-        pretrained.trainable = False
-
-        for layer in pretrained.layers:
-            layer.trainable = False
-
-        pretrained = pretrained(input_img)
-
-        finetuned = tf.keras.layers.Flatten()(pretrained)
-        finetuned = tf.keras.layers.Dense(512, activation="relu")(finetuned)
-        finetuned = tf.keras.layers.Dropout(0.2)(finetuned)
-        finetuned = tf.keras.layers.Dense(128, activation="relu")(finetuned)
-        finetuned = tf.keras.layers.Dropout(0.2)(finetuned)
-        finetuned = tf.keras.layers.Dense(32, activation="relu")(finetuned)
-        finetuned = tf.keras.layers.Dropout(0.2)(finetuned)
-        finetuned = tf.keras.layers.Dense(len(self.classes), activation="softmax")(finetuned)
-
-        model = tf.keras.models.Model(inputs=input_img, outputs=finetuned)
-
-        model.compile()
-
-        return model
+        raise NotImplementedError