oxocr 0.1.1__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (89) hide show
  1. oxocr/__init__.py +135 -0
  2. oxocr/cli.py +555 -0
  3. oxocr/config.py +421 -0
  4. oxocr/config_data/ocr_corrections.json +10 -0
  5. oxocr/errors.py +117 -0
  6. oxocr/facade.py +396 -0
  7. oxocr/models/__init__.py +1 -0
  8. oxocr/models/anaksor_ocr/__init__.py +4 -0
  9. oxocr/models/anaksor_ocr/detector.py +143 -0
  10. oxocr/models/anaksor_ocr/geometry.py +152 -0
  11. oxocr/models/anaksor_ocr/longline.py +79 -0
  12. oxocr/models/anaksor_ocr/pipeline.py +174 -0
  13. oxocr/models/anaksor_ocr/recognizer.py +132 -0
  14. oxocr/models/anaksor_ocr/specs.py +87 -0
  15. oxocr/models/anaksor_ocr/text.py +181 -0
  16. oxocr/models/base.py +353 -0
  17. oxocr/models/duguang_ocr/__init__.py +4 -0
  18. oxocr/models/duguang_ocr/detector.py +156 -0
  19. oxocr/models/duguang_ocr/pipeline.py +150 -0
  20. oxocr/models/duguang_ocr/recognizer.py +143 -0
  21. oxocr/models/duguang_ocr/seglink.py +374 -0
  22. oxocr/models/duguang_ocr/specs.py +159 -0
  23. oxocr/models/indra_ocr/__init__.py +5 -0
  24. oxocr/models/indra_ocr/detector.py +107 -0
  25. oxocr/models/indra_ocr/lm.py +248 -0
  26. oxocr/models/indra_ocr/mathrec.py +116 -0
  27. oxocr/models/indra_ocr/orient.py +178 -0
  28. oxocr/models/indra_ocr/pipeline.py +202 -0
  29. oxocr/models/indra_ocr/preproc.py +198 -0
  30. oxocr/models/indra_ocr/recognizer.py +130 -0
  31. oxocr/models/indra_ocr/rtl.py +106 -0
  32. oxocr/models/indra_ocr/specs.py +208 -0
  33. oxocr/models/indra_ocr/vocab.py +47 -0
  34. oxocr/models/languages.py +148 -0
  35. oxocr/models/openthai_ocr/__init__.py +4 -0
  36. oxocr/models/openthai_ocr/detector.py +116 -0
  37. oxocr/models/openthai_ocr/geometry.py +136 -0
  38. oxocr/models/openthai_ocr/order.py +175 -0
  39. oxocr/models/openthai_ocr/pipeline.py +318 -0
  40. oxocr/models/openthai_ocr/recognizer.py +120 -0
  41. oxocr/models/openthai_ocr/specs.py +89 -0
  42. oxocr/models/openthai_ocr/text.py +91 -0
  43. oxocr/models/pix2text/__init__.py +1 -0
  44. oxocr/models/pix2text/pipeline.py +163 -0
  45. oxocr/models/pix2text/specs.py +114 -0
  46. oxocr/models/pix2text/tokenizer.py +54 -0
  47. oxocr/models/rapid_latex_ocr/__init__.py +6 -0
  48. oxocr/models/rapid_latex_ocr/pipeline.py +171 -0
  49. oxocr/models/rapid_latex_ocr/preproc.py +95 -0
  50. oxocr/models/rapid_latex_ocr/specs.py +64 -0
  51. oxocr/models/rapid_ocr/__init__.py +1 -0
  52. oxocr/models/rapid_ocr/classifier.py +66 -0
  53. oxocr/models/rapid_ocr/detector.py +177 -0
  54. oxocr/models/rapid_ocr/pipeline.py +138 -0
  55. oxocr/models/rapid_ocr/recognizer.py +108 -0
  56. oxocr/models/rapid_ocr/specs.py +421 -0
  57. oxocr/pipeline.py +154 -0
  58. oxocr/postprocess/__init__.py +1 -0
  59. oxocr/postprocess/assemblers.py +124 -0
  60. oxocr/postprocess/base.py +328 -0
  61. oxocr/postprocess/corrections.py +198 -0
  62. oxocr/postprocess/ghost.py +77 -0
  63. oxocr/postprocess/grouping.py +36 -0
  64. oxocr/postprocess/joiners.py +162 -0
  65. oxocr/postprocess/normalize.py +58 -0
  66. oxocr/postprocess/render.py +43 -0
  67. oxocr/preprocess/__init__.py +1 -0
  68. oxocr/preprocess/base.py +208 -0
  69. oxocr/preprocess/dark_to_light.py +135 -0
  70. oxocr/preprocess/pad_border.py +73 -0
  71. oxocr/preprocess/theme.py +73 -0
  72. oxocr/registry/__init__.py +1 -0
  73. oxocr/registry/catalog.py +24 -0
  74. oxocr/registry/download.py +163 -0
  75. oxocr/registry/registry.py +224 -0
  76. oxocr/runtime/__init__.py +1 -0
  77. oxocr/runtime/executor.py +178 -0
  78. oxocr/runtime/imageops.py +390 -0
  79. oxocr/runtime/images.py +84 -0
  80. oxocr/runtime/pool.py +196 -0
  81. oxocr/runtime/session.py +55 -0
  82. oxocr/runtime/worker.py +86 -0
  83. oxocr/types.py +459 -0
  84. oxocr/visual.py +295 -0
  85. oxocr-0.1.1.dist-info/METADATA +234 -0
  86. oxocr-0.1.1.dist-info/RECORD +89 -0
  87. oxocr-0.1.1.dist-info/WHEEL +4 -0
  88. oxocr-0.1.1.dist-info/entry_points.txt +2 -0
  89. oxocr-0.1.1.dist-info/licenses/LICENSE +21 -0
oxocr/__init__.py ADDED
@@ -0,0 +1,135 @@
1
+ """oxocr:模型无关的 Python OCR 包。
2
+
3
+ 公共门面。导入本包不触发 onnxruntime / 常驻池创建(首次使用才惰性加载)。
4
+ """
5
+ from __future__ import annotations
6
+
7
+ from oxocr.config import (
8
+ Backend,
9
+ InferOptions,
10
+ LayoutMode,
11
+ ModelOptions,
12
+ OcrConfig,
13
+ OutputFormat,
14
+ PostprocessConfig,
15
+ PreprocessConfig,
16
+ SubprocessOptions,
17
+ )
18
+ from oxocr.errors import (
19
+ ChecksumMismatchError,
20
+ InferenceError,
21
+ InvalidImageError,
22
+ MissingDependencyError,
23
+ ModelDownloadError,
24
+ ModelFileMissingError,
25
+ ModelNotFoundError,
26
+ OxocrError,
27
+ PluginNotFoundError,
28
+ StageError,
29
+ SubprocessError,
30
+ SubprocessTimeoutError,
31
+ UnknownParameterError,
32
+ )
33
+ from oxocr.facade import (
34
+ OcrEngine,
35
+ download_model,
36
+ get_model_spec,
37
+ list_models,
38
+ load_model,
39
+ loaded_models,
40
+ ocr,
41
+ readtext,
42
+ register_model_impl,
43
+ register_spec,
44
+ unload_all,
45
+ unload_model,
46
+ )
47
+ from oxocr.models.base import (
48
+ Capabilities,
49
+ InferRequest,
50
+ ModelFile,
51
+ ModelInfo,
52
+ ModelOutput,
53
+ ModelSpec,
54
+ OcrModel,
55
+ )
56
+ from oxocr.types import (
57
+ Box,
58
+ CharBox,
59
+ DebugInfo,
60
+ Document,
61
+ Elapse,
62
+ LoadedModelInfo,
63
+ OcrResult,
64
+ StageLog,
65
+ TextBlock,
66
+ TextItem,
67
+ TextLine,
68
+ ThemeResult,
69
+ ThemeType,
70
+ )
71
+
72
+ __version__ = "0.1.1"
73
+
74
+ __all__ = [
75
+ "__version__",
76
+ # 门面
77
+ "ocr",
78
+ "readtext",
79
+ "OcrEngine",
80
+ "list_models",
81
+ "get_model_spec",
82
+ "download_model",
83
+ "load_model",
84
+ "unload_model",
85
+ "unload_all",
86
+ "loaded_models",
87
+ "LoadedModelInfo",
88
+ "register_spec",
89
+ "register_model_impl",
90
+ # 数据类型
91
+ "Box",
92
+ "CharBox",
93
+ "TextItem",
94
+ "TextLine",
95
+ "TextBlock",
96
+ "Document",
97
+ "OcrResult",
98
+ "Elapse",
99
+ "DebugInfo",
100
+ "StageLog",
101
+ "ThemeType",
102
+ "ThemeResult",
103
+ # 模型契约
104
+ "OcrModel",
105
+ "ModelSpec",
106
+ "ModelFile",
107
+ "ModelInfo",
108
+ "Capabilities",
109
+ "InferRequest",
110
+ "ModelOutput",
111
+ # 配置
112
+ "OcrConfig",
113
+ "ModelOptions",
114
+ "InferOptions",
115
+ "PreprocessConfig",
116
+ "PostprocessConfig",
117
+ "SubprocessOptions",
118
+ "LayoutMode",
119
+ "OutputFormat",
120
+ "Backend",
121
+ # 异常
122
+ "OxocrError",
123
+ "ModelNotFoundError",
124
+ "ModelFileMissingError",
125
+ "ModelDownloadError",
126
+ "ChecksumMismatchError",
127
+ "MissingDependencyError",
128
+ "InvalidImageError",
129
+ "UnknownParameterError",
130
+ "InferenceError",
131
+ "SubprocessError",
132
+ "SubprocessTimeoutError",
133
+ "StageError",
134
+ "PluginNotFoundError",
135
+ ]
oxocr/cli.py ADDED
@@ -0,0 +1,555 @@
1
+ """命令行入口:Docker 风格子命令 oxocr <COMMAND> <ACTION> [OPTIONS]。
2
+
3
+ -h 默认英文,-h --lang zh 输出中文(与 -h 顺序无关)。
4
+ """
5
+ from __future__ import annotations
6
+
7
+ import argparse
8
+ import re
9
+ import sys
10
+ import unicodedata
11
+ from functools import partial
12
+ from pathlib import Path
13
+
14
+ PROG = "oxocr"
15
+
16
+ MODELS_SHORTCUTS = {"ls", "list", "inspect", "search", "download", "pull"}
17
+
18
+
19
+ def T(en: str, zh: str, lang: str) -> str:
20
+ return zh if lang == "zh" else en
21
+
22
+
23
+ class BilingualHelpFormatter(argparse.HelpFormatter):
24
+ def __init__(self, prog, lang="en", **kw):
25
+ self.lang = lang
26
+ super().__init__(prog, **kw)
27
+
28
+ def _format_usage(self, usage, actions, groups, prefix):
29
+ if prefix is None:
30
+ prefix = "用法: " if self.lang == "zh" else "usage: "
31
+ return super()._format_usage(usage, actions, groups, prefix)
32
+
33
+ def _fill_text(self, text, width, indent):
34
+ # 保留 epilog 示例的换行(默认 HelpFormatter 会折成一段)
35
+ return "".join(indent + line for line in text.splitlines(keepends=True))
36
+
37
+
38
+ def _detect_lang(argv: list[str]) -> str:
39
+ # action="help" 遇到 -h 立即打印退出,等不到 --lang 解析,只能预扫描;
40
+ # 非法值原样返回,由 main 统一报错(否则 -h 在前时会跳过校验)
41
+ for i, a in enumerate(argv):
42
+ if a == "--lang":
43
+ return argv[i + 1] if i + 1 < len(argv) else "en"
44
+ if a.startswith("--lang="):
45
+ return a.split("=", 1)[1]
46
+ return "en"
47
+
48
+
49
+ def _first_non_option(argv: list[str]) -> str | None:
50
+ i = 0
51
+ while i < len(argv):
52
+ a = argv[i]
53
+ if a == "--lang":
54
+ i += 2
55
+ continue
56
+ if a.startswith("--lang=") or a.startswith("-"):
57
+ i += 1
58
+ continue
59
+ return a
60
+ return None
61
+
62
+
63
+ def _normalize_argv(argv: list[str]) -> list[str]:
64
+ """顶层简写 ls / list / inspect / search → models 子命令。"""
65
+ first = _first_non_option(argv)
66
+ if first in MODELS_SHORTCUTS:
67
+ idx = argv.index(first)
68
+ return argv[:idx] + ["models"] + argv[idx:]
69
+ return argv
70
+
71
+
72
+ def _decorate(p: argparse.ArgumentParser, lang: str) -> argparse.ArgumentParser:
73
+ p.add_argument("-h", "--help", action="help",
74
+ help=T("show this help message and exit",
75
+ "显示此帮助信息并退出", lang))
76
+ p.add_argument("--lang", choices=["en", "zh"], default="en",
77
+ help=T("help language, used with -h (default: en)",
78
+ "帮助语言,与 -h 配合使用(默认 en)", lang))
79
+ if lang == "zh":
80
+ p._positionals.title = "位置参数"
81
+ p._optionals.title = "选项"
82
+ return p
83
+
84
+
85
+ def _new_parser(lang: str, **kwargs) -> argparse.ArgumentParser:
86
+ kwargs.setdefault("formatter_class", partial(BilingualHelpFormatter, lang=lang))
87
+ kwargs.setdefault("add_help", False)
88
+ return _decorate(argparse.ArgumentParser(**kwargs), lang)
89
+
90
+
91
+ def _new_subparser(sub, name: str, lang: str, **kwargs) -> argparse.ArgumentParser:
92
+ kwargs.setdefault("formatter_class", partial(BilingualHelpFormatter, lang=lang))
93
+ kwargs.setdefault("add_help", False)
94
+ return _decorate(sub.add_parser(name, **kwargs), lang)
95
+
96
+
97
+ def _add_global_options(sp: argparse.ArgumentParser, lang: str) -> None:
98
+ sp.add_argument("-D", "--model-dir", default=None, metavar="DIR",
99
+ help=T("model root directory (overrides OXOCR_MODEL_DIR)",
100
+ "模型根目录(覆盖 OXOCR_MODEL_DIR)", lang))
101
+ sp.add_argument("--debug", action="store_true",
102
+ help=T("enable debug logging", "输出调试日志", lang))
103
+
104
+
105
+ def _add_list_options(sp: argparse.ArgumentParser, lang: str) -> None:
106
+ sp.add_argument("-i", "--installed-only", action="store_true",
107
+ help=T("only list installed models",
108
+ "仅列出已安装模型", lang))
109
+ sp.add_argument("--sort", choices=["released", "name", "size"],
110
+ default="released", metavar="FIELD",
111
+ help=T("sort by field (default: %(default)s)",
112
+ "排序字段(默认 %(default)s)", lang))
113
+ sp.add_argument("-r", "--reverse", action="store_true",
114
+ help=T("reverse sort order", "反转排序方向", lang))
115
+
116
+
117
+ def _version_text() -> str:
118
+ from oxocr import __version__
119
+
120
+ return f"{PROG} {__version__}"
121
+
122
+
123
+ class _VersionAction(argparse.Action):
124
+ def __init__(self, option_strings, dest, **kwargs):
125
+ super().__init__(option_strings, dest, nargs=0, **kwargs)
126
+
127
+ def __call__(self, parser, namespace, values, option_string=None):
128
+ print(_version_text())
129
+ parser.exit()
130
+
131
+
132
+ def _build_parser(lang: str = "en") -> argparse.ArgumentParser:
133
+ epilog_en = (
134
+ "Examples:\n"
135
+ " oxocr run image.png\n"
136
+ " oxocr run image.png -m ppocrv6_small -o json\n"
137
+ " oxocr run image.png -M layout -o json -c\n"
138
+ " oxocr run image.png --vis out.png\n"
139
+ " oxocr models ls -i\n"
140
+ " oxocr ls -i\n"
141
+ " oxocr models inspect ppocrv6_small\n"
142
+ " oxocr models download ppocrv6_tiny\n"
143
+ " oxocr search zh -f lang\n"
144
+ " oxocr -h --lang zh # Chinese help\n"
145
+ )
146
+ epilog_zh = (
147
+ "示例:\n"
148
+ " oxocr run image.png\n"
149
+ " oxocr run image.png -m ppocrv6_small -o json\n"
150
+ " oxocr run image.png -M layout -o json -c\n"
151
+ " oxocr run image.png --vis out.png\n"
152
+ " oxocr models ls -i\n"
153
+ " oxocr ls -i\n"
154
+ " oxocr models inspect ppocrv6_small\n"
155
+ " oxocr models download ppocrv6_tiny\n"
156
+ " oxocr search zh -f lang\n"
157
+ " oxocr -h --lang zh # 中文帮助\n"
158
+ )
159
+
160
+ p = _new_parser(
161
+ lang,
162
+ prog=PROG,
163
+ description=T("oxocr command-line OCR tool",
164
+ "oxocr 命令行识别工具", lang),
165
+ epilog=epilog_zh if lang == "zh" else epilog_en,
166
+ )
167
+ p.add_argument("-v", "--version", action=_VersionAction,
168
+ help=T("show program's version number and exit",
169
+ "显示版本号并退出", lang))
170
+
171
+ sub = p.add_subparsers(dest="command", metavar="COMMAND")
172
+
173
+ # ---- run ----
174
+ run = _new_subparser(
175
+ sub, "run", lang,
176
+ help=T("recognize an image", "识别图像", lang),
177
+ description=T("Recognize an image and print the result",
178
+ "识别图像并输出结果", lang),
179
+ )
180
+ run.add_argument("image", help=T("path to the image", "图像路径", lang))
181
+ run.add_argument("-m", "--model", default="ppocrv6_small",
182
+ help=T("model name (default: %(default)s)",
183
+ "模型名(默认 %(default)s)", lang))
184
+ run.add_argument("-M", "--mode", choices=["raw", "line", "para", "layout"],
185
+ default="para",
186
+ help=T("layout mode (default: %(default)s)",
187
+ "排版模式(默认 %(default)s)", lang))
188
+ run.add_argument("-o", "--output", "--format", dest="output",
189
+ choices=["text", "json"], default="text",
190
+ help=T("output format (default: %(default)s)",
191
+ "输出格式(默认 %(default)s)", lang))
192
+ run.add_argument("-c", "--cls", action="store_true",
193
+ help=T("enable orientation classification",
194
+ "开启方向分类", lang))
195
+ run.add_argument("-s", "--text-score", type=float, default=None, metavar="FLOAT",
196
+ help=T("recognition confidence threshold",
197
+ "识别置信度阈值", lang))
198
+ run.add_argument("-t", "--threads", type=int, default=None, metavar="N",
199
+ help=T("number of inference threads", "推理线程数", lang))
200
+ run.add_argument("--vis", default=None, metavar="PATH",
201
+ help=T("save a visualization image to PATH (format by extension)",
202
+ "保存可视化图像到 PATH(格式按扩展名)", lang))
203
+ run.add_argument("--vis-mode", choices=["box_txt", "box", "ocr"],
204
+ default="box_txt", metavar="MODE",
205
+ help=T("visualization mode: box_txt (side-by-side boxes and "
206
+ "text), box (boxes only), ocr (boxes + text panel) "
207
+ "(default: %(default)s)",
208
+ "可视化模式:box_txt 双联图(框+文字)、box 仅画框、"
209
+ "ocr 画框+文字面板(默认 %(default)s)", lang))
210
+ run.add_argument("--font", default=None, metavar="PATH",
211
+ help=T("font file (ttf/ttc) for visualization text "
212
+ "(default: auto-detect system font)",
213
+ "可视化文字字体文件(ttf/ttc,默认自动探测系统字体)", lang))
214
+ _add_global_options(run, lang)
215
+ run.set_defaults(func=_cmd_run)
216
+
217
+ # ---- models ----
218
+ models = _new_subparser(
219
+ sub, "models", lang,
220
+ aliases=["model"],
221
+ help=T("manage models", "模型管理", lang),
222
+ description=T("Manage OCR models", "模型管理", lang),
223
+ )
224
+ msub = models.add_subparsers(dest="models_cmd", metavar="ACTION")
225
+
226
+ mlist = _new_subparser(
227
+ msub, "list", lang,
228
+ aliases=["ls"],
229
+ help=T("list available models", "列出可用模型", lang),
230
+ description=T("List available models", "列出可用模型", lang),
231
+ )
232
+ _add_list_options(mlist, lang)
233
+ _add_global_options(mlist, lang)
234
+ mlist.set_defaults(func=_cmd_models_list)
235
+
236
+ minfo = _new_subparser(
237
+ msub, "info", lang,
238
+ aliases=["inspect"],
239
+ help=T("show model information", "显示模型信息", lang),
240
+ description=T("Show model information", "显示模型信息", lang),
241
+ )
242
+ minfo.add_argument("name", help=T("model name", "模型名", lang))
243
+ _add_global_options(minfo, lang)
244
+ minfo.set_defaults(func=_cmd_models_info)
245
+
246
+ msearch = _new_subparser(
247
+ msub, "search", lang,
248
+ help=T("search models by name or language",
249
+ "按名称或语言搜索模型", lang),
250
+ description=T("Search models by name or language",
251
+ "按名称或语言搜索模型", lang),
252
+ )
253
+ msearch.add_argument("keyword", help=T("search keyword", "搜索关键词", lang))
254
+ msearch.add_argument("-f", "--field", choices=["name", "family", "lang", "all"],
255
+ default=None, metavar="FIELD",
256
+ help=T("fields to search: name, family, lang, all "
257
+ "(default: name+lang)",
258
+ "检索字段:name、family、lang、all"
259
+ "(默认 name+lang)", lang))
260
+ _add_list_options(msearch, lang)
261
+ _add_global_options(msearch, lang)
262
+ msearch.set_defaults(func=_cmd_models_search)
263
+
264
+ mdownload = _new_subparser(
265
+ msub, "download", lang,
266
+ aliases=["pull"],
267
+ help=T("download model weights", "下载模型权重", lang),
268
+ description=T("Download model weights (ModelScope first, HuggingFace fallback)",
269
+ "下载模型权重(ModelScope 优先,失败回退 HuggingFace)", lang),
270
+ )
271
+ mdownload.add_argument("name", help=T("model name", "模型名", lang))
272
+ mdownload.add_argument("--source", choices=["auto", "modelscope", "huggingface"],
273
+ default=None, metavar="SRC",
274
+ help=T("download source (default: auto, "
275
+ "or env OXOCR_DOWNLOAD_SOURCE)",
276
+ "下载源(默认 auto,可用 OXOCR_DOWNLOAD_SOURCE 设置)", lang))
277
+ mdownload.add_argument("--force", action="store_true",
278
+ help=T("re-download even if files exist",
279
+ "文件已存在也重新下载", lang))
280
+ _add_global_options(mdownload, lang)
281
+ mdownload.set_defaults(func=_cmd_models_download)
282
+
283
+ models.set_defaults(models_cmd="list", installed_only=False,
284
+ sort="released", reverse=False,
285
+ func=_cmd_models_list, model_dir=None, debug=False)
286
+
287
+ # ---- version ----
288
+ ver = _new_subparser(
289
+ sub, "version", lang,
290
+ help=T("show version", "显示版本", lang),
291
+ description=T("Show version", "显示版本", lang),
292
+ )
293
+ ver.set_defaults(func=_cmd_version)
294
+
295
+ return p
296
+
297
+
298
+ def _save_visualization(args, result) -> None:
299
+ from PIL import Image
300
+
301
+ from oxocr.visual import draw_boxes, draw_ocr, draw_items
302
+
303
+ boxes = [it.box for it in result.items]
304
+ txts = [it.text for it in result.items]
305
+ scores = [it.score for it in result.items]
306
+ if args.vis_mode == "box":
307
+ vis = draw_boxes(args.image, boxes, scores)
308
+ elif args.vis_mode == "ocr":
309
+ vis = draw_ocr(args.image, boxes, txts, scores, font_path=args.font)
310
+ else:
311
+ vis = draw_items(args.image, result.items, font_path=args.font)
312
+ Image.fromarray(vis[:, :, ::-1]).save(args.vis)
313
+
314
+
315
+ def _cmd_run(args) -> int:
316
+ import oxocr
317
+
318
+ overrides: dict = {}
319
+ if args.cls:
320
+ overrides["use_cls"] = True
321
+ if args.text_score is not None:
322
+ overrides["text_score"] = args.text_score
323
+ if args.threads is not None:
324
+ overrides["num_threads"] = args.threads
325
+
326
+ try:
327
+ result = oxocr.ocr(
328
+ args.image,
329
+ model=args.model,
330
+ mode=args.mode,
331
+ output=args.output,
332
+ model_dir=args.model_dir,
333
+ debug=args.debug,
334
+ **overrides,
335
+ )
336
+ except oxocr.OxocrError as e:
337
+ print(f"错误: {e}", file=sys.stderr)
338
+ return 1
339
+
340
+ # result.text 反映 --output 选定的渲染格式(text/json)
341
+ print(result.text)
342
+
343
+ if args.vis:
344
+ try:
345
+ _save_visualization(args, result)
346
+ except (oxocr.OxocrError, OSError, ValueError) as e:
347
+ print(f"错误: {e}", file=sys.stderr)
348
+ return 1
349
+ print(T(f"Visualization saved: {args.vis}",
350
+ f"可视化已保存: {args.vis}", args.lang), file=sys.stderr)
351
+ return 0
352
+
353
+
354
+ def _disp_width(text: str) -> int:
355
+ # 中文等全角字符在终端占 2 列,按显示宽度对齐
356
+ return sum(2 if unicodedata.east_asian_width(c) in "FW" else 1 for c in text)
357
+
358
+
359
+ # 长协议在表格中缩写显示,完整值见 models inspect
360
+ _LICENSE_SHORT = {"AI PUBS OPEN RAIL-M (modified)": "RAIL-M (modified)"}
361
+
362
+
363
+ def _pad(text: str, width: int, *, right: bool = False) -> str:
364
+ fill = " " * max(0, width - _disp_width(text))
365
+ return fill + text if right else text + fill
366
+
367
+
368
+ def _model_table_rows(args, zh: bool) -> list[dict]:
369
+ import oxocr
370
+
371
+ rows = []
372
+ for info in oxocr.list_models(installed_only=args.installed_only,
373
+ model_dir=args.model_dir):
374
+ spec = oxocr.get_model_spec(info.name)
375
+ # ASCII 标记:✓ 在 GBK 控制台/重定向下会抛 UnicodeEncodeError
376
+ mark = "[V]" if info.installed else "[ ]"
377
+ released = spec.raw_release_date.isoformat() if spec.raw_release_date else "-"
378
+ size = f"{spec.size_bytes / 1048576:.1f}" if spec.size_bytes else "-"
379
+ langs = spec.languages_chinese if zh else spec.languages
380
+ shown = ", ".join(langs[:3]) + (", ..." if len(langs) > 3 else "")
381
+ license = spec.license or "-"
382
+ rows.append({"mark": mark, "name": info.name,
383
+ "released": released, "size": size,
384
+ "license": _LICENSE_SHORT.get(license, license),
385
+ "langs": shown or "-",
386
+ "date": spec.raw_release_date, "bytes": spec.size_bytes,
387
+ "spec": spec})
388
+ return rows
389
+
390
+
391
+ def _sort_rows(rows: list[dict], field: str, reverse: bool) -> list[dict]:
392
+ # released/size 自然方向为降序(最新/最大在前),name 为升序
393
+ desc = field in ("released", "size")
394
+ if reverse:
395
+ desc = not desc
396
+ if field == "name":
397
+ return sorted(rows, key=lambda r: r["name"].lower(), reverse=desc)
398
+ attr = "date" if field == "released" else "bytes"
399
+ valued = [r for r in rows if r[attr]]
400
+ missing = [r for r in rows if not r[attr]]
401
+ # 两遍稳定排序:同值按名称升序兜底,不受 reverse 影响
402
+ valued.sort(key=lambda r: r["name"].lower())
403
+ valued.sort(key=lambda r: r[attr], reverse=desc)
404
+ missing.sort(key=lambda r: r["name"].lower())
405
+ return valued + missing
406
+
407
+
408
+ def _print_model_table(rows: list[dict], zh: bool) -> None:
409
+ headers = (("名称", "开源日期", "大小(MB)", "开源协议", "支持语言") if zh else
410
+ ("NAME", "RELEASED", "SIZE(MB)", "LICENSE", "LANGUAGES"))
411
+ cols = ("name", "released", "size", "license", "langs")
412
+ widths = [max([_disp_width(headers[i])] + [_disp_width(r[c]) for r in rows])
413
+ for i, c in enumerate(cols)]
414
+ gaps = (" ", " ", " ", " ", "")
415
+
416
+ def line(cells: tuple[str, ...]) -> str:
417
+ return "".join(_pad(c, widths[i], right=(i == 2)) + gaps[i]
418
+ for i, c in enumerate(cells)).rstrip()
419
+
420
+ print(" " + line(headers))
421
+ for r in rows:
422
+ print(r["mark"] + " " + line(tuple(r[c] for c in cols)))
423
+
424
+
425
+ def _cmd_models_list(args) -> int:
426
+ zh = args.lang == "zh"
427
+ rows = _sort_rows(_model_table_rows(args, zh), args.sort, args.reverse)
428
+ _print_model_table(rows, zh)
429
+ return 0
430
+
431
+
432
+ def _match_spec(spec, keyword: str, field: str | None) -> bool:
433
+ q = keyword.lower()
434
+ if field is None:
435
+ fields = ("name", "lang")
436
+ elif field == "all":
437
+ fields = ("name", "family", "lang")
438
+ else:
439
+ fields = (field,)
440
+ for f in fields:
441
+ if f == "name":
442
+ if q in spec.name.lower() or q in spec.name_chinese.lower():
443
+ return True
444
+ elif f == "family":
445
+ if q in spec.family.lower():
446
+ return True
447
+ # 语言代码前缀匹配(zh 命中 zh/zh_tw);中文名子串;英文名按词前缀,
448
+ # 避免 ar 误中 Devanagari、hi 误中 Chinese
449
+ elif any(c.lower().startswith(q) for c in spec.languages) or any(
450
+ q in n.lower() for n in spec.languages_chinese) or any(
451
+ w.startswith(q) for n in spec.languages_english
452
+ for w in re.split(r"[^a-z0-9]+", n.lower()) if w):
453
+ return True
454
+ return False
455
+
456
+
457
+ def _cmd_models_search(args) -> int:
458
+ zh = args.lang == "zh"
459
+ rows = [r for r in _model_table_rows(args, zh)
460
+ if _match_spec(r["spec"], args.keyword, args.field)]
461
+ if not rows:
462
+ print(T("no matching models", "未找到匹配的模型", args.lang), file=sys.stderr)
463
+ return 1
464
+ _print_model_table(_sort_rows(rows, args.sort, args.reverse), zh)
465
+ return 0
466
+
467
+
468
+ def _cmd_models_info(args) -> int:
469
+ import oxocr
470
+ from oxocr.registry.registry import resolve_model_dir
471
+
472
+ try:
473
+ spec = oxocr.get_model_spec(args.name)
474
+ except oxocr.ModelNotFoundError as e:
475
+ print(f"错误: {e}", file=sys.stderr)
476
+ return 1
477
+
478
+ model_dir = resolve_model_dir(args.model_dir)
479
+ info = next(m for m in oxocr.list_models(model_dir=model_dir)
480
+ if m.name == spec.name)
481
+ cap = spec.capabilities
482
+ flags = [k for k in ("detection", "recognition", "angle_classification",
483
+ "char_boxes", "own_preprocess", "own_postprocess")
484
+ if getattr(cap, k)]
485
+ arch = ", ".join(f"{k}={v}" for k, v in spec.model_architecture.items()) or "-"
486
+ urls = "\n" + "\n".join(f"{'':<18}{u}" for u in spec.urls) if spec.urls else "-"
487
+ rows = [
488
+ ("Name", spec.name),
489
+ ("Chinese name", spec.name_chinese or spec.name),
490
+ ("Family", spec.family),
491
+ ("Type", spec.model_type),
492
+ ("Languages", ", ".join(spec.languages) or "-"),
493
+ ("Languages (zh)", ", ".join(spec.languages_chinese) or "-"),
494
+ ("Author", spec.author or "-"),
495
+ ("Country", spec.author_country or "-"),
496
+ ("Release date", spec.release_date.isoformat() if spec.release_date else "-"),
497
+ ("License", spec.license or "-"),
498
+ ("Commercial use", spec.commercial_use),
499
+ ("Star rating", str(spec.star_rating)),
500
+ ("Size", f"{spec.size_bytes / 1048576:.1f} MB" if spec.size_bytes else "-"),
501
+ ("Architecture", arch),
502
+ ("Installed", "yes" if info.installed else "no"),
503
+ ("Downloadable", "yes" if info.downloadable else "no"),
504
+ ("Model dir", str(model_dir)),
505
+ ("Requires", ", ".join(spec.requires) or "-"),
506
+ ("Capabilities", f"{', '.join(flags) or '-'} (backend: {cap.backend})"),
507
+ ("URLs", urls),
508
+ ]
509
+ for label, value in rows:
510
+ print(f"{label + ':':<18}{value}")
511
+ print("Files:")
512
+ for f in spec.files:
513
+ mark = "V" if Path(info.paths[f.role]).is_file() else "X"
514
+ size = f" ({f.size / 1048576:.1f} MB)" if f.size else ""
515
+ print(f" [{mark}] {f.role:6s} {f.filename}{size}")
516
+ return 0
517
+
518
+
519
+ def _cmd_models_download(args) -> int:
520
+ import oxocr
521
+
522
+ try:
523
+ paths = oxocr.download_model(
524
+ args.name, model_dir=args.model_dir, force=args.force, source=args.source,
525
+ )
526
+ except oxocr.OxocrError as e:
527
+ print(f"错误: {e}", file=sys.stderr)
528
+ return 1
529
+ print(T(f"Model {args.name} downloaded:", f"模型 {args.name} 下载完成:", args.lang))
530
+ for role, path in paths.items():
531
+ print(f" [{role}] {path}")
532
+ return 0
533
+
534
+
535
+ def _cmd_version(_args) -> int:
536
+ print(_version_text())
537
+ return 0
538
+
539
+
540
+ def main(argv: list[str] | None = None) -> int:
541
+ argv = list(sys.argv[1:] if argv is None else argv)
542
+ lang = _detect_lang(argv)
543
+ argv = _normalize_argv(argv)
544
+ parser = _build_parser(lang=lang if lang in ("en", "zh") else "en")
545
+ if lang not in ("en", "zh"):
546
+ parser.error(f"argument --lang: invalid choice: {lang!r} (choose from 'en', 'zh')")
547
+ args = parser.parse_args(argv)
548
+ if not hasattr(args, "func"):
549
+ parser.print_help()
550
+ return 2
551
+ return args.func(args) or 0
552
+
553
+
554
+ if __name__ == "__main__":
555
+ raise SystemExit(main())