Upload 4 files

README.md

@@ -0,0 +1,55 @@
+---
+license: apache-2.0
+tags:
+- dataset
+- document-ocr
+- text-tabular
+---
+
+# Document OCR Text Tabular Data Notes
+
+## Dataset summary
+
+Preparation notes and schema examples for **Document OCR** tasks using **Text Tabular** data. Full source material is intentionally not bundled, so provenance and licensing remain explicit.
+
+## Included material
+
+- `dataset.py` — loading, cleaning, and split preparation code.
+- `dataset_infos.json` — schema and split metadata.
+- `metadata_sample.jsonl` — small, human-readable records for checking the schema.
+- `README.md` — data card and usage notes.
+
+## Processing choices
+
+| Stage | Setting |
+|---|---|
+| Storage format | npy sharded |
+| Preprocessing | minimal |
+| Augmentation | mixup cutmix |
+| Split strategy | stratified 90 10 |
+| Sampling | balanced |
+| Quality checks | strict |
+| Labeling | pseudo label |
+
+## Validation checklist
+
+Before using the prepared data, verify source licenses, duplicates across splits, missing values, label balance, and modality-specific corruption. Record the source version and every filtering rule so a later run can reproduce the same rows.
+
+## Intended use
+
+The repository is suitable for testing the data pipeline, adapting it to a documented source, and preparing controlled research splits. Release status: **metadata sample; full source data not bundled**. The sample is for schema inspection only and should not be reported as a full training corpus.
+
+## Risks and limitations
+
+The loader cannot guarantee that an external source is representative, correctly licensed, or free of sensitive information. Users remain responsible for source review, privacy checks, and bias analysis before training or redistribution.
+
+## Files
+
+- `dataset.py` — primary artifact
+- `README.md` — this documentation
+- `dataset_infos.json` — schema metadata
+- `metadata_sample.jsonl` — schema sample
+
+## License
+
+Released under **apache-2.0**. Review the source-data terms separately when this repository is used with external datasets.

dataset.py

@@ -0,0 +1,159 @@
+import json, os, hashlib, logging
+from pathlib import Path
+
+log = logging.getLogger(__name__)
+
+# --- real data source: document_ocr ---
+TV_DATASET = None
+HF_CANDIDATES = ['nielsr/funsd', 'aharley/rvl_cdip']
+IMAGE_FIELD = 'image'
+TEXT_FIELD = 'words'
+LABEL_FIELD = 'label'
+PROMPT_TEMPLATE = 'a photo of a {label}'
+DATASET_URL = 'https://guillaumejaume.github.io/FUNSD/'
+
+def fetch_real_samples(max_samples=5000, cache_dir='./_cache'):
+    # 本地没有数据时自动下载真实公开数据集: torchvision -> HuggingFace -> 手动说明
+    out = []
+    if TV_DATASET is not None:
+        try:
+            import torchvision
+            ctor = getattr(torchvision.datasets, TV_DATASET)
+            try:
+                ds = ctor(root=cache_dir, split='train', download=True)
+            except TypeError:
+                try:
+                    ds = ctor(root=cache_dir, train=True, download=True)
+                except TypeError:
+                    ds = ctor(root=cache_dir, download=True)
+            classes = getattr(ds, 'classes', None)
+            os.makedirs(os.path.join(cache_dir, 'tv'), exist_ok=True)
+            for i, item in enumerate(ds):
+                if len(out) >= max_samples:
+                    break
+                img, label = item[0], item[1]
+                name = classes[label] if classes else str(label)
+                p = os.path.join(cache_dir, 'tv', str(i) + '.png')
+                try:
+                    img.save(p)
+                except Exception:
+                    continue
+                out.append({'image': p, 'text': PROMPT_TEMPLATE.format(label=name)})
+            if out:
+                return out
+        except Exception as e:
+            print('torchvision load failed:', e)
+    for repo in HF_CANDIDATES:
+        try:
+            from datasets import load_dataset
+            try:
+                ds = load_dataset(repo, split='train', streaming=True)
+            except Exception:
+                ds = load_dataset(repo, split='train')
+            img_dir = os.path.join(cache_dir, 'hf_images')
+            os.makedirs(img_dir, exist_ok=True)
+            for i, ex in enumerate(ds):
+                if len(out) >= max_samples:
+                    break
+                txt = None
+                if TEXT_FIELD is not None and TEXT_FIELD in ex:
+                    v = ex[TEXT_FIELD]
+                    txt = v if isinstance(v, str) else ' '.join(map(str, v if isinstance(v, (list, tuple)) else [v]))
+                if txt is None and LABEL_FIELD in ex:
+                    txt = PROMPT_TEMPLATE.format(label=ex[LABEL_FIELD])
+                if txt is None:
+                    continue
+                if IMAGE_FIELD not in ex or ex[IMAGE_FIELD] is None:
+                    continue
+                p = os.path.join(img_dir, str(i) + '.jpg')
+                try:
+                    ex[IMAGE_FIELD].convert('RGB').save(p)
+                except Exception:
+                    continue
+                out.append({'image': p, 'text': txt})
+            if out:
+                return out
+        except Exception as e:
+            print('HF load failed for', repo, ':', e)
+    print('Automatic download failed. Please get the data manually from:')
+    print('   ' + DATASET_URL)
+    return out
+
+
+def load(data_dir):
+    
+    files = list(Path(data_dir).glob('*.jsonl'))
+    if not files:
+        files = list(Path(data_dir).glob('*.json'))
+    samples = []
+    for f in files:
+        with open(f, encoding="utf-8") as fp:
+            if f.suffix == '.json':
+                d = json.load(fp)
+                samples.extend(d if isinstance(d, list) else [d])
+            else:
+                samples.extend(json.loads(l) for l in fp if l.strip())
+    return samples
+
+def filter_quality(samples, min_score=0.5):
+    
+    results = []
+    for s in samples:
+        text = s.get("document_ocr", s.get("text", ""))
+        if len(text.split()) >= 3:
+            results.append(s)
+    return results
+
+def dedup(samples):
+    seen = set()
+    out = []
+    for s in samples:
+        fp = s.get("image", s.get("audio", ""))
+        if fp and os.path.exists(fp):
+            h = hashlib.md5(open(fp, 'rb').read()).hexdigest()
+            if h in seen:
+                continue
+            seen.add(h)
+        out.append(s)
+    return out
+
+def preprocess(samples, out_dir, img_size=224):
+    # 预处理
+    os.makedirs(out_dir, exist_ok=True)
+    processed = []
+    for s in samples:
+        item = {}
+        if "image" in s:
+            try:
+                from PIL import Image as IM
+                img = IM.open(s["image"]).convert("RGB")
+                img = img.resize((img_size, img_size))
+                p = os.path.join(out_dir, os.path.basename(s["image"]))
+                img.save(p, "JPEG", quality=95)
+                item["image"] = p
+            except Exception:
+                continue
+        text = s.get("document_ocr", s.get("text", ""))
+        item["text"] = text
+        item["domain"] = "document_ocr"
+        processed.append(item)
+    return processed
+
+def save_jsonl(data, path):
+    with open(path, 'w', encoding="utf-8") as f:
+        for d in data:
+            f.write(json.dumps(d, ensure_ascii=False) + '\n')
+
+def main():
+    import sys
+    data_dir = sys.argv[1] if len(sys.argv) > 1 else './data'
+    out = sys.argv[2] if len(sys.argv) > 2 else './output'
+    samples = load(data_dir) or fetch_real_samples()
+    samples = filter_quality(samples)
+    samples = dedup(samples)
+    result = preprocess(samples, out)
+    save_jsonl(result, os.path.join(out, 'dataset.jsonl'))
+    print(f'Done: {len(result)} samples')
+
+if __name__ == '__main__':
+    main()

dataset_infos.json

@@ -0,0 +1,26 @@
+{
+  "description": "Metadata sample and preparation pipeline for document_ocr / text_tabular data",
+  "features": {
+    "id": {
+      "dtype": "string"
+    },
+    "description": {
+      "dtype": "string"
+    },
+    "label": {
+      "dtype": "int64"
+    },
+    "domain": {
+      "dtype": "string"
+    },
+    "modality": {
+      "dtype": "string"
+    }
+  },
+  "splits": {
+    "sample": {
+      "num_examples": 24
+    }
+  },
+  "release_status": "metadata sample; full source data not bundled"
+}

metadata_sample.jsonl

@@ -0,0 +1,24 @@
+{"id": "sample-001", "description": "schema validation example", "label": 0, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-002", "description": "quality-control example", "label": 1, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-003", "description": "split inspection example", "label": 2, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-004", "description": "label review example", "label": 3, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-005", "description": "deduplication example", "label": 0, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-006", "description": "format conversion example", "label": 1, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-007", "description": "schema validation example", "label": 2, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-008", "description": "quality-control example", "label": 3, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-009", "description": "split inspection example", "label": 0, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-010", "description": "label review example", "label": 1, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-011", "description": "deduplication example", "label": 2, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-012", "description": "format conversion example", "label": 3, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-013", "description": "schema validation example", "label": 0, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-014", "description": "quality-control example", "label": 1, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-015", "description": "split inspection example", "label": 2, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-016", "description": "label review example", "label": 3, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-017", "description": "deduplication example", "label": 0, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-018", "description": "format conversion example", "label": 1, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-019", "description": "schema validation example", "label": 2, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-020", "description": "quality-control example", "label": 3, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-021", "description": "split inspection example", "label": 0, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-022", "description": "label review example", "label": 1, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-023", "description": "deduplication example", "label": 2, "domain": "document_ocr", "modality": "text_tabular"}
+{"id": "sample-024", "description": "format conversion example", "label": 3, "domain": "document_ocr", "modality": "text_tabular"}