图重复了
Browse files- ocr/gendata.py +16 -6
ocr/gendata.py
CHANGED
|
@@ -29,6 +29,7 @@ with open(pth_txt, 'r', encoding='utf-8') as f:
|
|
| 29 |
|
| 30 |
|
| 31 |
dic = {}
|
|
|
|
| 32 |
num = 0
|
| 33 |
for PageId in PageIds:
|
| 34 |
print(PageId)
|
|
@@ -41,11 +42,18 @@ for PageId in PageIds:
|
|
| 41 |
assert len(images) > 0
|
| 42 |
pth_img = images[0]
|
| 43 |
pth_img = str( Path(pth_img).resolve() )
|
| 44 |
-
img = cv2.imread(pth_img)
|
| 45 |
-
if len(img.shape) != 3: # 转彩图
|
| 46 |
-
img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
|
| 47 |
-
img_color = img.copy()
|
| 48 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 49 |
|
| 50 |
txt = fp.read()
|
| 51 |
data = json.loads(txt)
|
|
@@ -77,8 +85,10 @@ for PageId in PageIds:
|
|
| 77 |
print(charText, end='')
|
| 78 |
print("\n\n\n")
|
| 79 |
|
| 80 |
-
|
| 81 |
-
|
|
|
|
|
|
|
| 82 |
|
| 83 |
|
| 84 |
pass
|
|
|
|
| 29 |
|
| 30 |
|
| 31 |
dic = {}
|
| 32 |
+
dic_img = {}
|
| 33 |
num = 0
|
| 34 |
for PageId in PageIds:
|
| 35 |
print(PageId)
|
|
|
|
| 42 |
assert len(images) > 0
|
| 43 |
pth_img = images[0]
|
| 44 |
pth_img = str( Path(pth_img).resolve() )
|
|
|
|
|
|
|
|
|
|
|
|
|
| 45 |
|
| 46 |
+
new_img = False
|
| 47 |
+
if PageId not in dic_img:
|
| 48 |
+
new_img = True
|
| 49 |
+
img = cv2.imread(pth_img)
|
| 50 |
+
if len(img.shape) != 3: # 转彩图
|
| 51 |
+
img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
|
| 52 |
+
img_color = img.copy()
|
| 53 |
+
dic_img[PageId] = { "img_color": img_color, "img": img }
|
| 54 |
+
else:
|
| 55 |
+
img = dic_img[PageId]["img"]
|
| 56 |
+
img_color = dic_img[PageId]["img_color"]
|
| 57 |
|
| 58 |
txt = fp.read()
|
| 59 |
data = json.loads(txt)
|
|
|
|
| 85 |
print(charText, end='')
|
| 86 |
print("\n\n\n")
|
| 87 |
|
| 88 |
+
|
| 89 |
+
if new_img:
|
| 90 |
+
cv2.imwrite(f"./out2/{num:04d}_{image_id}.webp", img_color)
|
| 91 |
+
num += 1
|
| 92 |
|
| 93 |
|
| 94 |
pass
|