feat(ocr): integrate dedicated Vietnamese PP-OCRv5 model and streamline dropdown

.gitattributes

@@ -324,3 +324,4 @@ ocr-studio/data/output/rec_cham_inference_v24/inference.pdmodel filter=lfs diff=
 ocr-studio/data/output/rec_cham_inference_v24/inference.pdiparams.info filter=lfs diff=lfs merge=lfs -text
 
 ocr-studio/data/output/ch_PP-OCRv4_det_cham_infer/inference.pdiparams filter=lfs diff=lfs merge=lfs -text
+ocr-studio/data/output/rec_vietnamese_infer/inference.pdiparams filter=lfs diff=lfs merge=lfs -text

ocr-studio/app.py

@@ -194,7 +194,7 @@ def get_ocr_model(version):
     if version in ocr_models:
         return ocr_models[version]
         
-    if version in ['vi', 'en', 'fr', 'japan']:
+    if version in ['en', 'fr', 'japan']:
         ocr_model = OfficialPaddleOCRWrapper(lang=version)
         ocr_models[version] = ocr_model
         return ocr_model
@@ -209,14 +209,40 @@ def get_ocr_model(version):
     from tools.infer.predict_rec import TextRecognizer
     import tools.infer.utility as utility
     
-    model_dir = os.path.join(PROJECT_ROOT, "data", "output", f"rec_cham_inference_{version}")
-    dict_path = os.path.join(PROJECT_ROOT, "data", f"cham_dict_{version}.txt")
-    
-    if not os.path.exists(model_dir):
-        # Fallback to Golden Baseline v23 if version-specific folder is missing
-        model_dir = os.path.join(PROJECT_ROOT, "data", "output", "rec_cham_inference_v23")
-    if not os.path.exists(dict_path):
-        dict_path = os.path.join(PROJECT_ROOT, "data", "cham_dict_v23.txt")
+    if version == 'vi':
+        model_dir = os.path.join(PROJECT_ROOT, "data", "output", "rec_vietnamese_infer")
+        dict_path = os.path.join(PROJECT_ROOT, "data", "vietnamese_dict.txt")
+        model_ready = os.path.exists(os.path.join(model_dir, "inference.pdiparams")) and os.path.exists(dict_path)
+        if not model_ready:
+            try:
+                print(f"📥 Downloading dedicated Vietnamese OCR model to {model_dir}...")
+                os.makedirs(model_dir, exist_ok=True)
+                import urllib.request
+                base_url = "https://huggingface.co/anbeo2004/paddleocr-v5-rec-vi-custom/resolve/main/"
+                dl_files = {
+                    "custom_dict.txt": dict_path,
+                    "inference.json": os.path.join(model_dir, "inference.json"),
+                    "inference.yml": os.path.join(model_dir, "inference.yml"),
+                    "inference.pdiparams": os.path.join(model_dir, "inference.pdiparams")
+                }
+                for fname, ftarget in dl_files.items():
+                    if not os.path.exists(ftarget):
+                        urllib.request.urlretrieve(base_url + fname, ftarget)
+                print("✅ Downloaded Vietnamese OCR model successfully.")
+            except Exception as e:
+                print(f"⚠️ Failed to auto-download Vietnamese model: {e}")
+                ocr_model = OfficialPaddleOCRWrapper(lang='vi')
+                ocr_models[version] = ocr_model
+                return ocr_model
+    else:
+        model_dir = os.path.join(PROJECT_ROOT, "data", "output", f"rec_cham_inference_{version}")
+        dict_path = os.path.join(PROJECT_ROOT, "data", f"cham_dict_{version}.txt")
+        
+        if not os.path.exists(model_dir):
+            # Fallback to Golden Baseline v23 if version-specific folder is missing
+            model_dir = os.path.join(PROJECT_ROOT, "data", "output", "rec_cham_inference_v23")
+        if not os.path.exists(dict_path):
+            dict_path = os.path.join(PROJECT_ROOT, "data", "cham_dict_v23.txt")
         
     sys_argv_backup = sys.argv
     sys.argv = [sys.argv[0]]

ocr-studio/data/output/rec_vietnamese_infer/inference.pdiparams

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb929c47e349f528c6ff7e00d67b101bf7bb7654dffb5d18ac4156550841ebd3
+size 75601159

ocr-studio/data/output/rec_vietnamese_infer/inference.yml

@@ -0,0 +1,267 @@
+Global:
+  model_name: PP-OCRv5_server_rec
+Hpi:
+  backend_configs:
+    paddle_infer:
+      trt_dynamic_shapes: &id001
+        x:
+        - - 1
+          - 3
+          - 48
+          - 160
+        - - 1
+          - 3
+          - 48
+          - 320
+        - - 8
+          - 3
+          - 48
+          - 3200
+    tensorrt:
+      dynamic_shapes: *id001
+PreProcess:
+  transform_ops:
+  - DecodeImage:
+      channel_first: false
+      img_mode: BGR
+  - MultiLabelEncode:
+      gtc_encode: NRTRLabelEncode
+  - RecResizeImg:
+      image_shape:
+      - 3
+      - 48
+      - 512
+  - KeepKeys:
+      keep_keys:
+      - image
+      - label_ctc
+      - label_gtc
+      - length
+      - valid_ratio
+PostProcess:
+  name: CTCLabelDecode
+  character_dict:
+  - '!'
+  - '"'
+  - '%'
+  - '&'
+  - ''''
+  - (
+  - )
+  - ','
+  - '-'
+  - .
+  - /
+  - '0'
+  - '1'
+  - '2'
+  - '3'
+  - '4'
+  - '5'
+  - '6'
+  - '7'
+  - '8'
+  - '9'
+  - ':'
+  - ;
+  - '?'
+  - A
+  - B
+  - C
+  - D
+  - E
+  - F
+  - G
+  - H
+  - I
+  - J
+  - K
+  - L
+  - M
+  - N
+  - O
+  - P
+  - Q
+  - R
+  - S
+  - T
+  - U
+  - V
+  - W
+  - X
+  - Y
+  - Z
+  - '['
+  - ']'
+  - _
+  - a
+  - b
+  - c
+  - d
+  - e
+  - f
+  - g
+  - h
+  - i
+  - j
+  - k
+  - l
+  - m
+  - n
+  - o
+  - p
+  - q
+  - r
+  - s
+  - t
+  - u
+  - v
+  - w
+  - x
+  - y
+  - z
+  - ²
+  - À
+  - Á
+  - È
+  - É
+  - Ê
+  - Ì
+  - Í
+  - Ð
+  - Ò
+  - Ó
+  - Ô
+  - Õ
+  - Ö
+  - Ù
+  - Ú
+  - Ü
+  - Ý
+  - à
+  - á
+  - â
+  - ã
+  - è
+  - é
+  - ê
+  - ì
+  - í
+  - ð
+  - ò
+  - ó
+  - ô
+  - õ
+  - ö
+  - ù
+  - ú
+  - ü
+  - ý
+  - Ā
+  - ā
+  - Ă
+  - ă
+  - Đ
+  - đ
+  - Ĩ
+  - ĩ
+  - Ō
+  - ō
+  - Ũ
+  - ũ
+  - Ū
+  - ū
+  - Ơ
+  - ơ
+  - Ư
+  - ư
+  - Ạ
+  - ạ
+  - Ả
+  - ả
+  - Ấ
+  - ấ
+  - Ầ
+  - ầ
+  - Ẩ
+  - ẩ
+  - Ẫ
+  - ẫ
+  - Ậ
+  - ậ
+  - Ắ
+  - ắ
+  - Ằ
+  - ằ
+  - Ẳ
+  - ẳ
+  - Ẵ
+  - ẵ
+  - Ặ
+  - ặ
+  - Ẹ
+  - ẹ
+  - Ẻ
+  - ẻ
+  - Ẽ
+  - ẽ
+  - Ế
+  - ế
+  - Ề
+  - ề
+  - Ể
+  - ể
+  - Ễ
+  - ễ
+  - Ệ
+  - ệ
+  - Ỉ
+  - ỉ
+  - Ị
+  - ị
+  - Ọ
+  - ọ
+  - Ỏ
+  - ỏ
+  - Ố
+  - ố
+  - Ồ
+  - ồ
+  - Ổ
+  - ổ
+  - Ỗ
+  - ỗ
+  - Ộ
+  - ộ
+  - Ớ
+  - ớ
+  - Ờ
+  - ờ
+  - Ở
+  - ở
+  - Ỡ
+  - ỡ
+  - Ợ
+  - ợ
+  - Ụ
+  - ụ
+  - Ủ
+  - ủ
+  - Ứ
+  - ứ
+  - Ừ
+  - ừ
+  - Ử
+  - ử
+  - Ữ
+  - ữ
+  - Ự
+  - ự
+  - Ỳ
+  - ỳ
+  - Ỵ
+  - ỵ
+  - Ỷ
+  - ỷ
+  - Ỹ
+  - ỹ

ocr-studio/data/vietnamese_dict.txt

@@ -0,0 +1,224 @@
+!
+"
+%
+&
+'
+(
+)
+,
+-
+.
+/
+0
+1
+2
+3
+4
+5
+6
+7
+8
+9
+:
+;
+?
+A
+B
+C
+D
+E
+F
+G
+H
+I
+J
+K
+L
+M
+N
+O
+P
+Q
+R
+S
+T
+U
+V
+W
+X
+Y
+Z
+[
+]
+_
+a
+b
+c
+d
+e
+f
+g
+h
+i
+j
+k
+l
+m
+n
+o
+p
+q
+r
+s
+t
+u
+v
+w
+x
+y
+z
+²
+À
+Á
+È
+É
+Ê
+Ì
+Í
+Ð
+Ò
+Ó
+Ô
+Õ
+Ö
+Ù
+Ú
+Ü
+Ý
+à
+á
+â
+ã
+è
+é
+ê
+ì
+í
+ð
+ò
+ó
+ô
+õ
+ö
+ù
+ú
+ü
+ý
+Ā
+ā
+Ă
+ă
+Đ
+đ
+Ĩ
+ĩ
+Ō
+ō
+Ũ
+ũ
+Ū
+ū
+Ơ
+ơ
+Ư
+ư
+Ạ
+ạ
+Ả
+ả
+Ấ
+ấ
+Ầ
+ầ
+Ẩ
+ẩ
+Ẫ
+ẫ
+Ậ
+ậ
+Ắ
+ắ
+Ằ
+ằ
+Ẳ
+ẳ
+Ẵ
+ẵ
+Ặ
+ặ
+Ẹ
+ẹ
+Ẻ
+ẻ
+Ẽ
+ẽ
+Ế
+ế
+Ề
+ề
+Ể
+ể
+Ễ
+ễ
+Ệ
+ệ
+Ỉ
+ỉ
+Ị
+ị
+Ọ
+ọ
+Ỏ
+ỏ
+Ố
+ố
+Ồ
+ồ
+Ổ
+ổ
+Ỗ
+ỗ
+Ộ
+ộ
+Ớ
+ớ
+Ờ
+ờ
+Ở
+ở
+Ỡ
+ỡ
+Ợ
+ợ
+Ụ
+ụ
+Ủ
+ủ
+Ứ
+ứ
+Ừ
+ừ
+Ử
+ử
+Ữ
+ữ
+Ự
+ự
+Ỳ
+ỳ
+Ỵ
+ỵ
+Ỷ
+ỷ
+Ỹ
+ỹ

ocr-studio/index.html

@@ -1254,19 +1254,13 @@
             <div class="model-picker-wrap" title="Select neural OCR recognition architecture">
                 <span class="model-picker-label">Model:</span>
                 <select id="model-select" class="model-select-compact" onchange="updateModelBadge()">
-                    <optgroup label="Paleographic Cham Models (Custom Fine-Tuned)">
-                        <option value="v24" selected>Version 24 (Validated Baseline — Verse-Numeral & Noise-Resilient SVTR)</option>
-                        <option value="v23">Version 23 (Golden Baseline — Visual Order SVTR)</option>
-                        <option value="v22">Version 22 (Legacy Fallback Baseline)</option>
+                    <optgroup label="Paleographic Cham Models">
+                        <option value="v24" selected>Cham V24 (Recommended — Logical Order)</option>
+                        <option value="v23">Cham V23 (Visual Order Baseline)</option>
                     </optgroup>
-                    <optgroup label="Multilingual Baselines (PaddleOCR PP-OCRv4)">
-                        <option value="vi">Vietnamese (Latin)</option>
-                        <option value="en">English (Latin)</option>
-                        <option value="fr">French (Latin)</option>
-                        <option value="japan">Japanese (CJK)</option>
-                    </optgroup>
-                    <optgroup label="Autonomous Routing">
-                        <option value="auto">Adaptive Dual-Engine (Cham & Vietnamese)</option>
+                    <optgroup label="Bilingual & Modern Text">
+                        <option value="auto">Auto Dual-Engine (Cham & Vietnamese)</option>
+                        <option value="vi">Vietnamese (Fine-Tuned PP-OCRv5)</option>
                     </optgroup>
                 </select>
             </div>