Delete README.md

README.md

@@ -1,50 +0,0 @@
-# Image editing checklist dataset
-
-This download contains 27,779 records and 37,379 unique image files.
-The 8 independent `.tar.gz` archives preserve the `images/...` paths
-referenced by `data.json`. Shared source images are stored only once.
-
-## Download and extract
-
-Download `data.json`, every `images-*.tar.gz` archive, and `SHA256SUMS` into
-one directory. Each archive can be extracted independently into that same
-directory. All archives are needed to use every record:
-
-```bash
-for archive in images-*.tar.gz; do
-  tar -xzf "$archive"
-done
-```
-
-After extraction, the directory contains `data.json` and `images/`.
-Resolve both image fields relative to this directory.
-
-## Fields
-
-- `instruction`: the editing instruction.
-- `source_image`: the relative path to the original image.
-- `edited_image`: the relative path to the edited image.
-- `checklist`: questions, dimensions, and binary answers (1 = satisfied,
-  0 = not satisfied).
-
-`data.json` and all image bytes are preserved without rewriting or resizing.
-The archive file size limit is in decimal GB (1 GB = 1,000,000,000 bytes).
-Gzip level 1 is used; these archives are not parts of a split gzip stream.
-
-## Integrity
-
-The packaging script writes `SHA256SUMS` only after all archives are complete.
-Use `pack_images.py --dst-root /path/to/this/directory --verify` for checksum
-and archive-content verification without extracting any files.
-On Linux, `sha256sum -c SHA256SUMS` performs checksum verification only;
-on macOS, use `shasum -a 256 -c SHA256SUMS`.
-
-## Distribution notes
-
-These are ordinary download archives, not WebDataset shards. The full JSON
-references images across archives. This layout does not configure the Hugging
-Face Dataset Viewer to display paired images automatically.
-
-Before public release, complete the dataset name, source/annotation provenance,
-citation and applicable license in this card. No license is assigned by this
-packaging utility.