Commit 30015313 authored by dmMaze's avatar dmMaze
Browse files

set internal ndarray image channel order to RGB

parent 72caedb6

modules/inpaint/base.py

+6 −6
Original line number Diff line number Diff line
@@ -69,9 +69,9 @@ class InpainterBase(BaseModule): 
                    non_text_region = np.where(non_text_msk > 0)

                    non_text_px = img[non_text_region]

                    average_bg_color = np.median(non_text_px, axis=0)

                    std_bgr = np.std(non_text_px - average_bg_color, axis=0)

                    std_max = np.max(std_bgr)

                    inpaint_thresh = 7 if np.std(std_bgr) > 1 else 10

                    std_rgb = np.std(non_text_px - average_bg_color, axis=0)

                    std_max = np.max(std_rgb)

                    inpaint_thresh = 7 if np.std(std_rgb) > 1 else 10

                    if std_max < inpaint_thresh:

                        img = img.copy()

                        img[np.where(ballon_msk > 0)] = average_bg_color
@@ -92,9 +92,9 @@ class InpainterBase(BaseModule): 
                        non_text_region = np.where(non_text_msk > 0)

                        non_text_px = im[non_text_region]

                        average_bg_color = np.median(non_text_px, axis=0)

                        std_bgr = np.std(non_text_px - average_bg_color, axis=0)

                        std_max = np.max(std_bgr)

                        inpaint_thresh = 7 if np.std(std_bgr) > 1 else 10

                        std_rgb = np.std(non_text_px - average_bg_color, axis=0)

                        std_max = np.max(std_rgb)

                        inpaint_thresh = 7 if np.std(std_rgb) > 1 else 10

                        if std_max < inpaint_thresh:

                            need_inpaint = False

                            im[np.where(ballon_msk > 0)] = average_bg_color

modules/ocr/mit32px.py

+2 −3
Original line number Diff line number Diff line
@@ -602,10 +602,9 @@ class OCR32pxModel: 
                    seq.append(ch)

                    

                textblk.text.append(''.join(seq))

                # manga-image-translator ocr extract bgr instead of rgb

                textblk.update_font_colors(

                    [fb, fg, fr],

                    [bb, bg, br]

                    [fr, fg, fb],

                    [br, bg, bb]

                )

            chunck_idx += N

modules/ocr/mit48px_ctc.py

+2 −2
Original line number Diff line number Diff line
@@ -464,8 +464,8 @@ class OCR48pxCTC: 
                    continue

                textblk.text.append(''.join(cur_texts))

                textblk.update_font_colors(

                    [int(total_fb()), int(total_fg()), int(total_fr())],

                    [int(total_bb()), int(total_bg()), int(total_br())]

                    [int(total_fr()), int(total_fg()), int(total_fb())],

                    [int(total_br()), int(total_bg()), int(total_bb())]

                )

            chunck_idx += N

modules/ocr/ocr_lens_proto.py

+1 −1
Original line number Diff line number Diff line
@@ -471,7 +471,7 @@ class OCRLensAPI_exp(OCRBase): 


        full_text = ""

        try:

            pil_img = Image.fromarray(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))

            pil_img = Image.fromarray(img)

            processed_bytes, width, height = _preprocess_image_for_lens(pil_img)



            if not processed_bytes:

modules/ocr/ocr_oneocr.py

+3 −14
Original line number Diff line number Diff line
@@ -150,18 +150,7 @@ class OcrEngine: 
    def recognize_pil(self, image: PilImage.Image):

        if image.mode != 'RGBA':

            image = image.convert('RGBA')

        try:

            r, g, b, a = image.split()

            # Try BGRA reorder needed by some libs

            bgra_image = PilImage.merge('RGBA', (b, g, r, a))

        except ValueError:

            if self.logger:

                self.logger.warning(

                    "Could not split RGBA, trying via RGB->BGRA.")

            rgb = image.convert('RGB')

            r, g, b = rgb.split()

            bgra_image = PilImage.merge('RGB', (b, g, r)).convert('RGBA')

        return self._process_image(cols=bgra_image.width, rows=bgra_image.height, step=bgra_image.width*4, data=bgra_image.tobytes())

        return self._process_image(cols=image.width, rows=image.height, step=image.width*4, data=image.tobytes())



    def _process_image(self, cols, rows, step, data):

        dp = ctypes.cast(data, c_ubyte_p) if not isinstance(
@@ -364,9 +353,9 @@ class OCROneAPI(OCRBase): 
            if len(img_to_process.shape) == 2:

                img_rgb = cv2.cvtColor(img_to_process, cv2.COLOR_GRAY2RGB)

            elif img_to_process.shape[2] == 3:

                img_rgb = cv2.cvtColor(img_to_process, cv2.COLOR_BGR2RGB)

                img_rgb = img_to_process

            elif img_to_process.shape[2] == 4:

                img_rgb = cv2.cvtColor(img_to_process, cv2.COLOR_BGRA2RGB)

                img_rgb = cv2.cvtColor(img_to_process, cv2.COLOR_RGBA2RGB)

            else:

                raise ValueError(

                    f"Unsupported channels: {img_to_process.shape[2]}")