layout changes

backend/README.md

@@ -0,0 +1,15 @@
+# Aoriil Backend
+
+# Project Description
+
+# Installation
+
+# Example Usage
+
+# Dev setup
+
+# Change Log
+
+# License
+
+# Contributing

backend/example.py

@@ -0,0 +1,19 @@
+"""
+
+"""
+
+import cv2
+
+import src.core.imagetranslation as imagetranslation
+
+
+image_path = "images/japanese4.png"
+image = cv2.imread(image_path)
+
+
+translated_image = imagetranslation.translate(image)
+
+cv2.imshow("Original image ", image)
+cv2.imshow("Translated Image ", translated_image)
+cv2.waitKey()
+cv2.destroyAllWindows()

backend/examples/image_conversion_example.py

@@ -0,0 +1,19 @@
+"""
+
+"""
+
+import cv2
+
+import backend.src.core.imagetranslation as imagetranslation
+
+
+image_path = "images/japanese4.png"
+image = cv2.imread(image_path)
+
+
+translated_image = imagetranslation.translate(image)
+
+cv2.imshow("Original image ", image)
+cv2.imshow("Translated Image ", translated_image)
+cv2.waitKey()
+cv2.destroyAllWindows()

backend/src/core/imagetranslation.py

@@ -0,0 +1,234 @@
+"""
+This module, imagetranslation, contains the main function that will translate
+the image to the destination language. 
+
+TODO Currently, the only supported destination lang is English. We want to be
+able to support multiple destination languages in future
+
+TODO Currently, the only supported images are of the following source languages
+    - japanese
+    - korean
+    - mandarin
+
+We want to be able to support multiple languages in future.
+
+"""
+
+import cv2
+import numpy as np
+import pytesseract
+
+# Modules used
+from src.modules.coreimage import calc_max_gap_dist, replace_image_section
+from src.modules.fontdetection import calculate_font_size
+from src.modules.textdetection import get_sections
+from src.modules.textinsertion import manage_text
+from src.modules.textsection import (
+    should_merge_sections,
+    textsection,
+    parenttextsection,
+)
+
+
+# Translation APIs
+from src.modules.translators.unofficial_google_trans import (
+    translate_to_destination_lang,
+)
+
+
+# Helper utilities used
+from src.utilities.helpers.imageHelpers import (
+    crop_image,
+    convert_cv2_to_pil,
+    convert_pil_to_cv2,
+    calculate_box_width,
+    calculate_box_height,
+)
+from src.utilities.helpers.stringHelpers import remove_trailing_whitespace
+
+
+# TODO: Support for  multiple other source languages and destination languages
+# When this is added, make sure to remove default value from src_lang but no
+# need to remove default value from dest_lang
+# TODO: Need to support robust error handling incase of some failures. As of
+# this commit, there virtually exists no error handling...
+def translate(
+    image: np.array,
+    src_lang: str = "japanese",
+    dest_lang: str = "english",
+    translation_service: str = "python-googletrans",
+    show_borders: bool = False,
+    save_image: bool = False,
+) -> np.array:
+    """
+    The image to translate its content from source lang to destination lang
+
+    Parameters
+    ----------
+    image : np.array
+        The image to translate its content
+
+    src_lang : str, optional
+        The image's primary language to translate from. Default is Japanese
+
+    dest_lang : str, optional
+        The language to translate all image's content to. Default is English
+
+    translation_service : str, optional
+        The translation API to use. Default is python.googletrans which is free
+
+    show_borders : bool, optional
+        Draws borders around detected text regions if True. Default is False
+
+    save_image : bool, optional
+        Saves the translated images to the ./results folder. If the folder does
+        not exists, it will create one at the root folder. Default is False
+
+    Returns
+    -------
+    np.array
+        The translated image
+    """
+
+    # General config sections
+    # TODO: Validate the source language before we do any processing
+
+    # TODO: Validate the destination language before we do any processing
+
+    # TODO: Validate if the provided translation service is supported
+
+    # TODO: Validate if proper API credentials are provided in .configs file
+
+    # Start Processing Section
+    pytesseract_config = r"--oem 3 --psm 5 -l jpn_vert"
+
+    # TODO: if language is part of vertical text, we MUST rotate it:
+    image = cv2.rotate(image, cv2.ROTATE_90_CLOCKWISE)
+
+    easyocr_sections = get_sections(image, source_lang=src_lang)
+
+    text_sections = []
+    for x_min, x_max, y_min, y_max in easyocr_sections:
+        text_sections.append(textsection("", [x_min, x_max, y_min, y_max], 12))
+
+    MAXIMUM_GAP_DIST = calc_max_gap_dist(image=image)
+    merged_sections = []
+
+    for section in text_sections:
+        if len(merged_sections) == 0:
+            merged_sections.append(parenttextsection(section))
+            continue
+
+        is_section_merged = False
+        for msection in merged_sections:
+            if should_merge_sections(
+                msection, section, MAXIMUM_GAP_DIST, image
+            ):
+                is_section_merged = True
+                msection.add_section(section)
+                break
+        if not is_section_merged:
+            merged_sections.append(parenttextsection(section))
+
+    destination_image = image.copy()
+    for section in merged_sections:
+        cropped_section = crop_image(
+            destination_image, section.start_pos, section.end_pos
+        )
+
+        # TODO: Change the way we mask the original text to remove it from
+        # the image. Right now, once we have the merged section, we simply
+        # cover it entirely by white and add a black background on the outside
+        # so that when we pass it to cv2.inpaint, it "attempts" to inpaint it
+        # However, based of examples, this doesnt look as good and thus needs
+        # to be revamped....
+        border_size = 1
+        mask = np.full(
+            (
+                section.height - border_size * 2,
+                section.width - border_size * 2,
+                1,
+            ),
+            255,
+            np.uint8,
+        )  # white img
+        mask = cv2.copyMakeBorder(
+            mask,
+            top=border_size,
+            bottom=border_size,
+            left=border_size,
+            right=border_size,
+            borderType=cv2.BORDER_CONSTANT,
+            value=(0, 0, 0),
+        )
+        processed_cropped_section = cv2.inpaint(
+            cropped_section, mask, 7, cv2.INPAINT_NS
+        )
+
+        replace_image_section(
+            destination_image, processed_cropped_section, section.start_pos
+        )
+
+    # TODO: if language is part of vertical text, we MUST rotate it:
+    rotate_img_height, rotate_img_width, _ = destination_image.shape
+    destination_image = cv2.rotate(
+        destination_image, cv2.ROTATE_90_COUNTERCLOCKWISE
+    )
+    image = cv2.rotate(image, cv2.ROTATE_90_COUNTERCLOCKWISE)
+
+    destination_image_pil = convert_cv2_to_pil(destination_image)
+    for section in merged_sections:
+        # TODO: if language is part of vertical text, we MUST rotate it
+        start_pos = [
+            section.start_pos[1],
+            rotate_img_width - section.end_pos[0],
+        ]
+        end_pos = [section.end_pos[1], rotate_img_width - section.start_pos[0]]
+
+        cropped_section = crop_image(image, start_pos, end_pos)
+
+        # TODO: if language is part of vertical text, we must use diff config
+        current_text = pytesseract.image_to_string(
+            cropped_section, config=pytesseract_config
+        )
+
+        current_text = remove_trailing_whitespace(current_text).replace(
+            " ", ""
+        )
+
+        if current_text == "":
+            continue
+
+        scale = round(rotate_img_height / 1000)
+        if scale == 0:
+            scale = 1
+
+        text_font_size = calculate_font_size(
+            current_text, start_pos, end_pos, scale=scale
+        )
+
+        translated_text = ""
+        if translation_service == "python-googletrans":
+            translated_text = translate_to_destination_lang(
+                current_text, src_lang
+            )
+        box = [
+            start_pos,
+            [end_pos[0], start_pos[1]],
+            end_pos,
+            [start_pos[0], end_pos[1]],
+        ]
+        box_width = calculate_box_width(box)
+        box_height = calculate_box_height(box)
+
+        destination_image_pil = manage_text(
+            destination_image_pil,
+            translated_text,
+            start_pos,
+            box_width,
+            box_height,
+            text_font_size,
+        )
+
+    destination_image = convert_pil_to_cv2(destination_image_pil)
+    return destination_image

backend/modules/fontdetection.py → backend/src/modules/fontdetection.py

@@ -22,7 +22,7 @@
 
 # TODO: We want to specify paths to font file outside of this file, in a
 # different location. This is an low priority enhancement
-DEFAULT_FONT_PATH = "utilities/fonts/Wild-Words-Roman.ttf"
+DEFAULT_FONT_PATH = "src/utilities/fonts/Wild-Words-Roman.ttf"
 
 
 # TODO: What should be the default font size used on most mangas? Is 12
@@ -122,7 +122,7 @@ def calculate_font_size(
     font_path = DEFAULT_FONT_PATH
 
     if font_name == "wild-words":
-        font_path = "utilities/fonts/Wild-Words-Roman.ttf"
+        font_path = "src/utilities/fonts/Wild-Words-Roman.ttf"
 
     left = MINIMUM_FONT_SIZE * scale
     right = MAXIMUM_FONT_SIZE * scale

backend/modules/textdetection.py → backend/src/modules/textdetection.py

@@ -9,7 +9,7 @@
 import numpy as np
 import easyocr
 
-import utilities.helpers.imageHelpers as imgutil
+from src.utilities.helpers.imageHelpers import unpack_box
 
 
 # TODO: Should this be initalized here or on the main file? Find better place
@@ -46,7 +46,7 @@ def draw_text_borders(image: np.array, borders_list: list) -> np.array:
     border_pixel = 1
 
     for box, _, _ in borders_list:
-        (tl, _, br, _) = imgutil.unpack_box(box)
+        (tl, _, br, _) = unpack_box(box)
 
         cv2.rectangle(
             image, tl, br, color=border_color, thickness=border_pixel

backend/modules/textinsertion.py → backend/src/modules/textinsertion.py

@@ -309,7 +309,7 @@ def manage_text(
     box_width: int,
     box_height: int,
     font_size: int = 12,
-    font_path: str = "utilities/fonts/Wild-Words-Roman.ttf",
+    font_path: str = "src/utilities/fonts/Wild-Words-Roman.ttf",
     padding: int = 2,
     line_height: int = LINE_HEIGHT_DEFAULT_VAL,
     vertical_text: bool = False,

backend/modules/textsection.py → backend/src/modules/textsection.py

@@ -9,8 +9,7 @@
 
 """
 
-import utilities.helpers.imageHelpers as imgutil
-
+from src.utilities.helpers.imageHelpers import unpack_box
 import cv2
 import math
 import numpy as np
@@ -603,9 +602,7 @@ def is_section_overlap(section1: list, section2: list) -> bool:
         True if a section overlaps another, False otherwise
     """
 
-    section1_tl, section1_tr, section1_br, section1_bl = imgutil.unpack_box(
-        section1
-    )
+    section1_tl, section1_tr, section1_br, section1_bl = unpack_box(section1)
 
     is_inside_section2 = (
         is_point_in_rectangle(section1_tl, section2)
@@ -617,9 +614,7 @@ def is_section_overlap(section1: list, section2: list) -> bool:
     if is_inside_section2:
         return True
 
-    section2_tl, section2_tr, section2_br, section2_bl = imgutil.unpack_box(
-        section2
-    )
+    section2_tl, section2_tr, section2_br, section2_bl = unpack_box(section2)
 
     is_inside_section1 = (
         is_point_in_rectangle(section2_tl, section1)
@@ -652,7 +647,7 @@ def is_point_in_rectangle(point: list, rectangle: list) -> bool:
         True if a point lies in the rectangle, False otherwise
     """
 
-    (tl, _, br, _) = imgutil.unpack_box(rectangle)
+    (tl, _, br, _) = unpack_box(rectangle)
     x1, y1, x2, y2 = tl[0], tl[1], br[0], br[1]
     x, y = point[0], point[1]