Merge pull request #65 from RapidAI/fix_table_cls_preprocess

fix: fix table cls preprocess

Author: Joker1212

lineless_table_rec/utils_table_recover.py

@@ -289,7 +289,7 @@ def gather_ocr_list_by_row(ocr_list: List[Any], thehold: float = 0.2) -> List[An
                 cur[0], next[0], axis="y", threhold=thehold
             )
             if c_idx:
-                dis = max(next_box[0] - cur_box[1], 0)
+                dis = max(next_box[0] - cur_box[0], 0)
                 blank_str = int(dis / threshold) * " "
                 cur[1] = cur[1] + blank_str + next[1]
                 xmin = min(cur_box[0], next_box[0])

table_cls/main.py

@@ -5,7 +5,7 @@
 import numpy as np
 from PIL import Image
 
-from .utils import InputType, LoadImage, OrtInferSession
+from .utils import InputType, LoadImage, OrtInferSession, resize_and_center_crop
 
 cur_dir = Path(__file__).resolve().parent
 q_cls_model_path = cur_dir / "models" / "table_cls.onnx"
@@ -70,7 +70,7 @@ def __init__(self, model_path):
 
     def preprocess(self, img):
         img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
-        img = cv2.resize(img, (640, 640))
+        img = resize_and_center_crop(img, 640)
         img = np.array(img, dtype=np.float32) / 255
         img = img.transpose(2, 0, 1)  # HWC to CHW
         img = np.expand_dims(img, axis=0)  # Add batch dimension, only one image

table_cls/utils.py

@@ -178,3 +178,37 @@ def cvt_four_to_three(img: np.ndarray) -> np.ndarray:
     def verify_exist(file_path: Union[str, Path]):
         if not Path(file_path).exists():
             raise LoadImageError(f"{file_path} does not exist.")
+
+
+def resize_and_center_crop(image, output_size=640):
+    """
+    将图片的最小边缩放到指定大小，并进行中心裁剪。
+
+    :param image: 输入的图片数组 (H, W, C)
+    :param output_size: 缩放和裁剪后的图片大小，默认为 640
+    :return: 处理后的图片数组 (output_size, output_size, C)
+    """
+    # 获取图片的高度和宽度
+    height, width = image.shape[:2]
+    # 计算缩放比例
+    if width < height:
+        new_width = output_size
+        new_height = int(output_size * height / width)
+    else:
+        new_width = int(output_size * width / height)
+        new_height = output_size
+
+    # 缩放图片
+    image_resize = cv2.resize(
+        image, (new_width, new_height), interpolation=cv2.INTER_LINEAR
+    )
+
+    # 计算中心裁剪的坐标
+    left = (new_width - output_size) // 2
+    top = (new_height - output_size) // 2
+    right = left + output_size
+    bottom = top + output_size
+
+    # # 中心裁剪
+    image_cropped = image_resize[top:bottom, left:right]
+    return image_cropped

wired_table_rec/utils_table_recover.py

@@ -383,7 +383,7 @@ def gather_ocr_list_by_row(ocr_list: List[Any], threhold: float = 0.2) -> List[A
                 cur[0], next[0], axis="y", threhold=threhold
             )
             if c_idx:
-                dis = max(next_box[0] - cur_box[1], 0)
+                dis = max(next_box[0] - cur_box[0], 0)
                 blank_str = int(dis / threshold) * " "
                 cur[1] = cur[1] + blank_str + next[1]
                 xmin = min(cur_box[0], next_box[0])