CoolFace
Datasetpublic

Helios1208/taobao-product-context

Taobao Product Context Private product-detail snapshots for agents that consume ordered product images and OCR-derived context. Dataset versions Dataset: 0.1.0 Schema: 1.0.0 Snapshot: 2026-08-31 Pipeline: generated from the pipeline Git commit recorded in each row Configurations products One row per strictly validated product snapshot. Each row contains source metadata, ordered relative image paths, image hashes and dimensions… See the full description on the dataset page: https://huggingface.co/datasets/Helios1208/taobao-product-context.

sourceHugging Faceupdated 24d agoView on Hugging Face
0likes43downloads
Dataset Card

Taobao Product Context

Private product-detail snapshots for agents that consume ordered product images and OCR-derived context.

Dataset versions

  • —Dataset: 0.1.0
  • —Schema: 1.0.0
  • —Snapshot: 2026-08-31
  • —Pipeline: generated from the pipeline Git commit recorded in each row

Configurations

products

One row per strictly validated product snapshot. Each row contains source metadata, ordered relative image paths, image hashes and dimensions, ordered ocr_pages, page-to-source-image provenance, and cleaned context_text.

failures

One row per planned product that did not meet the products acceptance gate. Failures are retained for audit and future resume; they are not silently dropped.

Acceptance gate

A product is published only when its source URL exists, collection metadata is complete, all expected image files are readable and contiguous, PaddleOCR-VL completed every page, the Agent context is non-empty, and the manifest maps every source image to an OCR page.

For snapshot 2026-08-31, 168 products passed this page-level contract. The dataset does not claim that individual text spans can be attributed to one source image when an OCR page contains multiple images.

Known limitations

  • —Historical product-level collection timestamps were not recorded. collected_at is null and collection_time_source is unknown for this snapshot.
  • —Category membership reflects the source workbook and may contain irrelevant search results. No semantic relevance label is claimed.
  • —OCR text may contain recognition errors. Historical IMAGE_START / IMAGE_END marker residue is removed during export.
  • —Exact duplicate images are retained and identified by SHA-256.
  • —Product pages can change or disappear after collection.

Access and usage

This dataset is private. Product images and merchant-authored text may be subject to platform terms and third-party rights. Do not make the repository public without a separate rights and policy review. Honor removal requests and avoid using the dataset to identify individuals or infer sensitive attributes.

Private access does not grant recipients permission to redistribute the images or merchant-authored text. No public redistribution license is asserted by this dataset card.

Excluded artifacts

Browser profiles, authentication state, API credentials, signed URLs, PaddleOCR job metadata, local absolute paths, logs, workbooks, raw OCR responses, and stitched OCR pages are not included.

Reproduction

Use src/export_hf_dataset.py to create the staging tree and src/validate_hf_dataset.py to verify it before upload. The source images and OCR roots remain read-only inputs.