validating single image predict for fashion_mnist - kamangir/bolt#692

image_classifier/__init__.py

@@ -1,5 +1,5 @@
 name = "image_classifier"
 
-version = "1.1.151"
+version = "1.1.156"
 
 description = "fashion-mnist + hugging-face + awesome-bash-cli"

image_classifier/__main__.py

@@ -1,4 +1,5 @@
 import argparse
+import cv2
 from . import *
 from .classes import *
 from .funcs import *
@@ -108,7 +109,7 @@ elif args.task == "predict":
             default=None,
         )
 
-        success = classifier.predict(
+        success, prediction = classifier.predict(
             test_images / 255.0,
             test_labels,
             args.output_path,
@@ -132,10 +133,25 @@ elif args.task == "predict_image":
         success, image = file.load_image(image_filename)
 
     if success:
-        success = classifier.predict(
+        image = cv2.resize(
+            image, (classifier.params["window_size"], classifier.params["window_size"])
+        )
+
+        if not classifier.params["color"]:
+            image = np.mean(image, axis=2)
+
+        image = np.expand_dims(image, axis=0)
+
+        success, prediction = classifier.predict(
             image / 255.0,
             output_path=args.output_path,
         )
+
+    if success:
+        index = np.argmax(prediction)
+        logger.info(
+            f"prediction: {classifier.class_names[index]} - {prediction[0][index]:.2f}"
+        )
 elif args.task == "preprocess":
     success = preprocess(
         args.output_path,

image_classifier/classes.py

@@ -102,12 +102,12 @@ class Image_Classifier(object):
         )
 
         if not output_path:
-            return True
+            return True, predictions
 
         if not file.save(
             f"{output_path}/image_classifier/predictions.pyndarray", predictions
         ):
-            return False
+            return False, predictions
 
         if test_labels is not None:
             from sklearn.metrics import confusion_matrix
@@ -126,7 +126,7 @@ class Image_Classifier(object):
             if not file.save(
                 f"{output_path}/image_classifier/model/confusion_matrix.pyndarray", cm
             ):
-                return False
+                return False, predictions
 
             if not graphics.render_confusion_matrix(
                 cm,
@@ -138,7 +138,7 @@ class Image_Classifier(object):
                 ],
                 footer=self.signature(prediction_time),
             ):
-                return False
+                return False, predictions
 
         if test_labels is not None:
             logger.info(
@@ -161,7 +161,7 @@ class Image_Classifier(object):
                 footer=self.signature(prediction_time),
                 title="distribution of test_labels",
             ):
-                return False
+                return False, predictions
 
         max_index = test_images.shape[0]
         if page_count != -1:
@@ -181,7 +181,7 @@ class Image_Classifier(object):
                 prediction_time,
             )
 
-        return True
+        return True, predictions
 
     def predict_frame(self, frame):
         prediction_time = time.time()