move cv into backend directory

.gitignore

@@ -1,11 +1,3 @@
 /.idea/
 /.vscode/
 /frontend/.vscode/
-/cv/testImages/kanti_img1.jpg
-/cv/testImages/kanti_img2.jpg
-/cv/testImages/kanti_img1.jpeg
-/cv/testImages/kanti_img2.jpeg
-/cv/testImages/kanti_telegram_compressed_1.jpg
-/cv/testImages/kanti_telegram_compressed_2.jpg
-/cv/__pycache__/
-/cv/Models/__pycache__/

README.md

@@ -21,16 +21,11 @@ We should be using virtual environments to not have problems with other versions
 
 1. Download Anaconda for Windows: https://www.anaconda.com/products/distribution#Downloads, add to PATH
 2. Run 'conda create --name virtualenv python=3.8'
-3. conda activate virtualenv
-4. pip install opencv-contrib-python
-5. pip3 install torch torchvision torchaudio
-6. pip install keras
-7. pip install tensorflow
-8. pip install imutils
+3. pip install -r requirements.txt
 
 To deactivate the virtual environment: 'conda deactivate'
 
 ## Use anaconda with vscode
 
 1. Ctrl+Shift+P -> Enter "Python: Select Interpreter"
-2. Select the virtualenv you just created.
+2. Select the virtualenv you just created.

backend/README.md

@@ -8,4 +8,4 @@
 
 ## Start Unit Testing
 1. Import all requirements using <code>pip install -r requirements.txt</code>
-2. Navigate to the backend/app directory and execute <code>pytest</code>
+2. Navigate to the backend/app directory and execute <code>pytest</code>

backend/app/app_cv.py

@@ -0,0 +1,42 @@
+import cv2
+import time
+from cv.DigitRecognizer import DigitRecognizer
+
+IMAGEDIR = "cv/testImages/"
+
+images = []
+
+images.append(cv2.imread(IMAGEDIR + "kanti_img1.jpeg"))
+images.append(cv2.imread(IMAGEDIR + "kanti_img2.jpeg"))
+images.append(cv2.imread(IMAGEDIR + "kanti_telegram_compressed_1.jpg"))
+images.append(cv2.imread(IMAGEDIR + "kanti_telegram_compressed_2.jpg"))
+images.append(cv2.imread(IMAGEDIR + "straight.jpg"))
+images.append(cv2.imread(IMAGEDIR + "perspective.jpg"))
+images.append(cv2.imread(IMAGEDIR + "crooked.jpg"))
+images.append(cv2.imread(IMAGEDIR + "lighting.jpg"))
+images.append(cv2.imread(IMAGEDIR + "mirror.jpg"))
+images.append(cv2.imread(IMAGEDIR + "multiple.jpg"))
+images.append(cv2.imread(IMAGEDIR + "rug.jpg"))
+images.append(cv2.imread(IMAGEDIR + "wavy.jpg"))
+images.append(cv2.imread(IMAGEDIR + "weird_bg.jpg"))
+images.append(cv2.imread(IMAGEDIR + "crunched.jpg"))
+
+
+recognizer = DigitRecognizer()
+
+cv_result = recognizer.recognize_digits_in_photo(images[0])
+
+#cv2.imshow('image', image)
+for index, image in enumerate(images):
+    start = time.time()
+    try:
+        cv_result = recognizer.recognize_digits_in_photo(image)
+        print(', '.join([str(exercise.score) for exercise in cv_result.exam.exercises]) + " | Total: " + str(cv_result.exam.score))
+
+        print("IMAGE " + str(index+1) + " PASSED")
+    except Exception as e:
+        print(e)
+        print("IMAGE " + str(index+1) + " DID NOT PASS")
+        pass
+    end = time.time()
+    print(end - start)

backend/app/core/cv_result.py

@@ -1,8 +1,7 @@
 class CVResult:
-    def __init__(self, candidate, exam, result_validated):
+    def __init__(self, candidate, exam):
         self.candidate = candidate
         self.exam = exam
-        self.result_validated = result_validated
 
 
 class Candidate:
@@ -12,12 +11,11 @@ def __init__(self, number, date_of_birth):
 
 
 class Exam:
-    def __init__(self, year, subject, total_score, total_score_confidence, exercises):
+    def __init__(self, year, subject, score, confidence, exercises):
         self.year = year
         self.subject = subject
-        self.total_score = total_score
-        #This is the confidence of the total score reached (how many points there are in the total cell)
-        self.total_score_confidence = total_score_confidence
+        self.score = score
+        self.confidence = confidence
         self.exercises = exercises
 
     def calc_exercises_score(self):

backend/app/core/scanner.py

@@ -2,11 +2,6 @@
 import datetime, os, uuid
 from fastapi import UploadFile
 from api.schema import BaseResponse, ExamFullResponse
-
-import sys
-
-sys.path.append(sys.path[0] + '/../../')
-sys.path.append(sys.path[0] + '/../../cv/')
 from cv.DigitRecognizer import DigitRecognizer
 
 from .admin import get_exam_full

backend/app/cv/.gitignore

@@ -0,0 +1,8 @@
+/cv/testImages/kanti_img1.jpg
+/cv/testImages/kanti_img2.jpg
+/cv/testImages/kanti_img1.jpeg
+/cv/testImages/kanti_img2.jpeg
+/cv/testImages/kanti_telegram_compressed_1.jpg
+/cv/testImages/kanti_telegram_compressed_2.jpg
+/cv/__pycache__/
+/cv/Models/__pycache__/

cv/DigitRecognizer.py → backend/app/cv/DigitRecognizer.py

@@ -1,15 +1,13 @@
 import cv2
 import base64
 import numpy as np
-from DocumentSegmentationCV import DocumentSegmentationCV
-from DocumentSegmentationCNN import DocumentSegmentationCNN
 from imutils import contours as imutils_contours
-from Models.mobilenet import MobileNet
-import Models.utils as utils
 
-import sys
-sys.path.append(sys.path[0] + '/..')
-from backend.app.core.cv_result import *
+from cv.Models.mobilenet import MobileNet
+import cv.Models.utils as utils
+from cv.DocumentSegmentationCV import DocumentSegmentationCV
+from cv.DocumentSegmentationCNN import DocumentSegmentationCNN
+import core.cv_result as cv_res
 
 class DigitRecognizer:
 
@@ -66,7 +64,7 @@ def recognize_digits_in_photo(self, photo):
 
         model = MobileNet()
         model.compile()
-        model.load_weights('./cv/Models/MobileNet.h5')
+        model.load_weights('cv/Models/MobileNet.h5')
         class_labels = ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9']
 
         exercises = []
@@ -93,19 +91,15 @@ def recognize_digits_in_photo(self, photo):
 
                 pred_class_label, pred_confidence = self.predict_handwritten_cell(result_cell, class_labels, model)
 
-                exercises.append(Exercise(index, pred_class_label, pred_confidence, "?"))
+                exercises.append(cv_res.Exercise(index, pred_class_label, pred_confidence, 0))  # TODO replace 0 with max_score
             elif(index % column_count != 0):
                 #print("Last Handwritten Cell, 'Total'")
 
                 total_score, total_score_confidence = self.predict_double_number(result_cell, class_labels, model)
 
-        exam = Exam("?", "?", total_score, total_score_confidence, exercises)
+        exam = cv_res.Exam(2023, "Mathematik", total_score, total_score_confidence, exercises)  # TODO replace year & subject
 
-        cv_res = CVResult(Candidate("?", "?"), exam=exam, result_validated=False)
-
-        if(exam.total_score == exam.calc_exercises_score()):
-            cv_res.result_validated = True
-        return cv_res
+        return cv_res.CVResult(cv_res.Candidate("NR-123", "2000-12-31"), exam=exam)  # TODO replace number & date_of_birth
 
 
     def predict_double_number(self, original_cell, class_labels, model):

cv/Models/base_model.py → backend/app/cv/Models/base_model.py

@@ -1,5 +1,5 @@
 from keras.utils import plot_model
-import Models.utils as utils
+from .utils import get_val_generator, get_train_generator
 
 class BaseModel(object):
     def __init__(self, model, optimizer, callbacks = None):
@@ -30,9 +30,9 @@ def fit_generator(self, training_data, validation_data, epochs, batch_size):
         x_train, y_train = training_data
         x_val, y_val = validation_data
 
-        train_datagen = utils.get_train_generator(x_train, y_train,
+        train_datagen = get_train_generator(x_train, y_train,
                                                   batch_size = batch_size)
-        val_datagen = utils.get_val_generator(x_val, y_val,
+        val_datagen = get_val_generator(x_val, y_val,
                                               batch_size = batch_size)
         hist = self.model.fit_generator(train_datagen,
                                         callbacks = self.callbacks,

cv/Models/mobilenet.py → backend/app/cv/Models/mobilenet.py

@@ -1,11 +1,10 @@
 from keras.callbacks import ReduceLROnPlateau
-from keras.applications import MobileNet as mobNet
 from keras.layers import (Input, Conv2D, BatchNormalization, ZeroPadding2D,
                           GlobalAveragePooling2D, Activation, Dense, DepthwiseConv2D)
 from keras.models import Model
 from keras import optimizers
-from Models.base_model import BaseModel
-from Models.train import train
+from .base_model import BaseModel
+from .train import train
 
 ALPHA = 1
 MODEL_NAME = f'MobileNet' # This should be modified when the model name changes.

cv/Models/train.py → backend/app/cv/Models/train.py

@@ -1,5 +1,5 @@
 import argparse
-import Models.utils as utils
+from .utils import load_mnist
 
 def get_argument_parser(model_name):
     '''
@@ -57,7 +57,7 @@ def train(model, model_name):
     # load all arguments
     args = get_argument_parser(model_name)
 
-    training_data, validation_data, test_data = utils.load_mnist()
+    training_data, validation_data, test_data = load_mnist()
     print(f'[data loaded]')
 
     # build and compile the model

cv/Models/utils.py → backend/app/cv/Models/utils.py

@@ -1,5 +1,4 @@
 import numpy as np
-from collections import defaultdict
 from keras.utils import np_utils
 from keras.datasets import mnist
 from keras.preprocessing.image import ImageDataGenerator

backend/app/requirements.txt

@@ -1,3 +1,4 @@
+# backend
 fastapi==0.95.0
 uvicorn==0.21.1
 peewee==3.16.0
@@ -6,4 +7,13 @@ pytest==7.2.2
 httpx==0.23.3
 asyncio==3.4.3
 nest-asyncio==1.5.1
-python-multipart==0.0.5
+python-multipart==0.0.5
+
+# backend/cv
+opencv-contrib-python==4.7.0.72
+torch==2.0.0
+torchvision==0.15.1
+torchaudio==2.0.1
+keras==2.12.0
+tensorflow==2.12.0
+imutils==0.5.4