guillemhs/manga-translator
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Added new styles

Browse Source
This commit is contained in:
Guillem Hernandez Sola
2026-04-21 21:45:46 +02:00
parent a5c81f4ff0
commit f00647e668
1 changed files with 18 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

35
manga-translator.py
View File

@@ -286,6 +286,7 @@ def split_wide_ocr_items(image_bgr, filtered_ocr):
return new_filtered, splits_made
def split_panel_box(image_bgr, bbox_xyxy, bubble_quads=None):
x1, y1, x2, y2 = bbox_xyxy
w = x2 - x1
@@ -367,20 +368,20 @@ def split_bubble_if_multiple_columns(indices, ocr, bid=None, use_aggressive_thre
med_h = float(np.median(hs)) if hs else 15.0
if use_aggressive_thresholds:
threshold1 = 80.0
threshold2 = med_h * 1.2
min_gap = 40.0
threshold1 = 60.0
threshold2 = med_h * 1.0
min_gap = 20.0
else:
threshold1 = 120.0
threshold2 = med_h * 3.0
min_gap = 60.0
threshold1 = 90.0
threshold2 = med_h * 1.5
min_gap = 25.0
if max_gap_size > threshold1 or (max_gap_size > threshold2 and max_gap_size > min_gap):
split_idx = max_gap_idx
left_indices = [item[0] for item in sorted_items[:split_idx]]
right_indices = [item[0] for item in sorted_items[split_idx:]]
if len(left_indices) < 2 or len(right_indices) < 2:
if len(left_indices) < 1 or len(right_indices) < 1:
return None
return left_indices, right_indices
@@ -413,8 +414,8 @@ def split_bubble_if_multiple_rows(indices, ocr, bid=None):
hs = [b[3] - b[1] for b in boxes]
med_h = float(np.median(hs)) if hs else 15.0
threshold = med_h * 2.5
min_gap = 40.0
threshold = med_h * 1.8
min_gap = 20.0
if max_gap_size > threshold and max_gap_size > min_gap:
split_idx = max_gap_idx
@@ -767,7 +768,7 @@ def group_tokens(ocr, image_shape, gap_px=18, bbox_padding=3):
centers = [quad_center(r[0]) for r in ocr]
hs = [max(1.0, b[3] - b[1]) for b in boxes]
med_h = float(np.median(hs)) if hs else 12.0
dist_thresh = max(20.0, med_h * 2.2)
dist_thresh = max(20.0, med_h * 1.8)
adaptive_gap_y = max(gap_px, med_h * 2.5)
@@ -793,7 +794,11 @@ def group_tokens(ocr, image_shape, gap_px=18, bbox_padding=3):
cx2, cy2 = centers[j]
is_vertically_aligned = abs(cx1 - cx2) < (med_h * 1.5)
if is_vertically_aligned and gap_y <= (med_h * 4.0):
if gap_x == 0 and gap_y <= (med_h * 3.5):
unite(i, j)
continue
if is_vertically_aligned and gap_y <= (med_h * 3.5):
unite(i, j)
continue
@@ -802,7 +807,7 @@ def group_tokens(ocr, image_shape, gap_px=18, bbox_padding=3):
continue
d = ((cx1 - cx2) ** 2 + (cy1 - cy2) ** 2) ** 0.5
if d <= dist_thresh and abs(cy1 - cy2) <= med_h * 3.0:
if d <= dist_thresh and abs(cy1 - cy2) <= med_h * 1.5:
unite(i, j)
groups = {}
@@ -1013,7 +1018,6 @@ def translate_manga_text(
print("⚠️ No text after filtering.")
return
# --- NEW: Split wide OCR items across column gaps ---
filtered, splits_made = split_wide_ocr_items(image, filtered)
if splits_made > 0:
print(f"✂️ Split {splits_made} wide OCR lines across column gaps.")
@@ -1030,7 +1034,6 @@ def translate_manga_text(
box = bubble_boxes[bid]
bubble_split = None
# 1. Panel border split
split_result = split_panel_box(image, box, bubble_quads=bubble_quads[bid])
if split_result:
box_left, box_right, split_x = split_result
@@ -1054,7 +1057,6 @@ def translate_manga_text(
bubble_split = (left_idxs, right_idxs)
splits_performed.append(f"BOX#{bid} ({len(left_idxs)} quads | {len(right_idxs)} quads)")
# 2. Check for vertical columns (left/right split)
if bubble_split is None:
col_split = split_bubble_if_multiple_columns(bubble_indices[bid], filtered, bid=bid)
if col_split:
@@ -1063,7 +1065,6 @@ def translate_manga_text(
bubble_split = (left_idxs, right_idxs)
splits_performed.append(f"BOX#{bid} (Vertical Column Split: {len(left_idxs)} | {len(right_idxs)} quads)")
# 3. Check for horizontal rows (top/bottom split)
if bubble_split is None:
row_split = split_bubble_if_multiple_rows(bubble_indices[bid], filtered, bid=bid)
if row_split:
@@ -1204,7 +1205,7 @@ def translate_manga_text(
if __name__ == "__main__":
translate_manga_text(
image_path="09.jpg",
image_path="004.png",
source_lang="english",
target_lang="ca",
confidence_threshold=0.05,