Start normalizing and preparing data for SVM

src/python/psic/classify/wash-over/visualize.py

@@ -5,29 +5,55 @@
 import cv2
 import numpy as np
 import pandas as pd
+from PIL import Image
+from sklearn.model_selection import train_test_split
 
-random.seed = 405
+from psic.resizer.generate import ResizeImages
+
+SEED = 405
+
+random.seed = SEED
 
 if getuser() == 'mattm':
     DRIVE_PATH = 'F:\\Shared drives\\P-Sick'
 else:
     DRIVE_PATH = 'mnt/Secondary/mcmoretz@uncg.edu/C-Sick'
 
 FINAL_TAGS_CSV = path.join(DRIVE_PATH, 'tag_csv/tagging_data.csv')
-SMALL_IMAGES_DIR = path.join(DRIVE_PATH, 'small/Florence/20180917a_jpgs/jpgs')
+SMALL_IMAGES_DIR = path.join(DRIVE_PATH, 'vsmall/5/Florence/20180917a_jpgs/jpgs')
+
+# Create new compressed images that are 5% of original image sizes using nearest neighbor selection
+ResizeImages.resize_all_images(path=path.join(DRIVE_PATH, 'data/Florence/20180917a_jpgs/jpgs'),
+                               output_path=SMALL_IMAGES_DIR,
+                               scale=0.05,
+                               img_filter=Image.NEAREST)
 
 data = pd.read_csv(FINAL_TAGS_CSV, usecols=['image_id', 'washover'])
 
 data['image'] = np.NaN
 print(data)
 
+X = list()  # The features of the data
+y = None  # The labels of the data
+
 for i, row in data.iterrows():
-    print(row.keys())
+    print('\rLoaded %s of %s images ' % (i, len(data)) + '.' * (i % 3), end='')
     image_path = path.join(SMALL_IMAGES_DIR, row['image_id'])
 
-    # Load grayscale versions of the small images
-    image = cv2.imread(image_path, 0)
-    row['image'] = image
+    # Load a 2d array of grayscale values
+    image: np.ndarray = cv2.imread(image_path, 0)
 
-data.drop(columns=['image_id'], inplace=True)
-print(data)
+    # String each row together to form a single 1d array of features
+    image = image.ravel()
+
+    # Create a row as a DataFrame with all the features as columns
+    # features: pd.DataFrame = pd.DataFrame(image.reshape(-1, len(image)))
+
+    X.append(list(image))
+
+print('\rLoaded all of the images!')
+print(pd.DataFrame(X, columns=range(len(X[0]))))
+
+
+# Split into test and training sets
+X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.33, random_state=SEED)