Organize CryoAtlas downloads and add searchable catalogs, result tables and density previews

Add archive lookup, SHA256SUMS and selective download helper. Preserve all scientific archives and original evaluation CSV bytes. Organize evaluation CSVs under benchmark/results and fixed split lists under metadata/splits. Add native Dataset Viewer configurations and documentation.

README.md

@@ -7,152 +7,170 @@ tags:
 - protein-structure
 - benchmark
 configs:
+- config_name: catalog
+  default: true
+  data_files:
+  - split: train
+    path: catalog/train.parquet
+  - split: test
+    path: catalog/test.parquet
+- config_name: benchmark_targets
+  data_files:
+  - split: test
+    path: catalog/benchmark_targets.parquet
+- config_name: benchmark_results
+  data_files:
+  - split: test
+    path: benchmark/results.parquet
+- config_name: examples
+  data_files:
+  - split: examples
+    path: previews/examples.parquet
+- config_name: downloads
+  data_files:
+  - split: archives
+    path: catalog/archives.parquet
 - config_name: modelangelo
   data_files:
   - split: test
-    path: Benchmark_outcome/ModelAngelo_results_618_full.csv
-  default: true
+    path: benchmark/results/ModelAngelo_results_618_full.csv
 - config_name: cryoatom
   data_files:
   - split: test
-    path: Benchmark_outcome/CryoAtom_results_618_full.csv
+    path: benchmark/results/CryoAtom_results_618_full.csv
 - config_name: e3_cryofold
   data_files:
   - split: test
-    path: Benchmark_outcome/E3-CryoFold_results_618_full.csv
+    path: benchmark/results/E3-CryoFold_results_618_full.csv
 - config_name: emprot
   data_files:
   - split: test
-    path: Benchmark_outcome/EMProt_results_618_full.csv
+    path: benchmark/results/EMProt_results_618_full.csv
 - config_name: af3
   data_files:
   - split: test
-    path: Benchmark_outcome/AF3_results_618_full.csv
+    path: benchmark/results/AF3_results_618_full.csv
 ---
 
-# CryoAtlas: A Large Curated Dataset and Unified Benchmark for Cryo-EM Atomic Model Building
+# CryoAtlas
+### A Large Curated Dataset and Unified Benchmark for Cryo-EM Atomic Model Building
 
-## Files currently hosted in this repository
+**17,809 entries · 16,657 training · 1,152 test · 618 benchmark targets**
 
-This repository hosts the complete CryoAtlas v1.0 release: all four training parts (68 independent archives; 16,657 entries), the complete test set (two archives containing the 618 benchmark targets and the 534 remaining test entries), benchmark resources, global and subset metadata, and five evaluation CSVs. Training archives are available under `train/part01/` through `train/part04/`, and test archives under `test/`. Subset metadata and file inventories are provided under `metadata/`, with archive checksums under `checksums/`. All 70 training and test archives have been verified against the frozen release using file sizes and SHA-256 checksums. The benchmark resource archive contains the predictions and examples described below.
+[Browse entries](https://huggingface.co/datasets/markfaa/CryoAtlas/viewer/catalog/train) · [Compare results](https://huggingface.co/datasets/markfaa/CryoAtlas/viewer/benchmark_results/test) · [Preview examples](https://huggingface.co/datasets/markfaa/CryoAtlas/viewer/examples/examples) · [Download guide](docs/downloads.md)
 
-- `resources/CryoAtlas_v1.0_benchmark_resources.tar.gz`: the complete benchmark resource package.
-- `Benchmark_outcome/`: individually accessible evaluation CSVs, configured as five Dataset Viewer subsets with 618 test targets each.
-- `metadata/`: global split and source metadata, plus the resource-file checksum inventory. Use the inventory from within the extracted resource folder.
-- `checksums/`: the resource archive's SHA256 file. From the repository root, verify the archive with `cd resources && shasum -a 256 -c ../checksums/CryoAtlas_v1.0_benchmark_resources.tar.gz.sha256`.
+CryoAtlas v1.0 pairs experimental cryo-EM density maps with atomic structures and protein sequences, using EMDB and PDB records available up to March 2026. Entries undergo map–model alignment, sequence mapping, and quality filtering to support research on automated protein atomic model building.
 
-## Overview
+The **complete release is available here**: four training parts, the full test set, and benchmark resources. The 70 training and test archives total **2.415 TB** (decimal units); their sizes and SHA-256 hashes have been verified against the frozen v1.0 release.
 
-CryoAtlas v1.0 provides curated experimental cryo-EM density maps, corresponding atomic structures, and protein sequences for research on automated protein atomic model building. The resource contains **17,809 map–model–sequence entries** derived from EMDB and PDB records available up to March 2026, following map–model alignment, sequence mapping, and quality filtering.
+![CryoAtlas split counts and reported map-resolution distributions](assets/dataset_overview.png)
 
-The fixed split comprises **16,657 training entries** and **1,152 test entries**. The paper's **618-target common benchmark is a subset of the test set**.
+## Browse before downloading
 
-## Release components
+The Dataset Viewer above displays lightweight tables and image previews. Select a configuration from its dropdown:
 
-The CryoAtlas v1.0 Package groups four training parts, the complete test set, and benchmark resources.
+| Configuration | Contents | Useful for |
+|---|---|---|
+| **`catalog`** — default | 16,657 training and 1,152 test rows | Search EMD/PDB IDs, resolution and titles; find the archive containing an entry |
+| **`benchmark_targets`** | The 618 common benchmark targets | Inspect the exact evaluation subset |
+| **`benchmark_results`** | 3,090 rows: 618 targets × five methods | Compare released metrics using the `emd_id` and `method` columns |
+| **`examples`** | Density projections for 10 example entries | Inspect sample maps without downloading a large archive |
+| **`downloads`** | 71 archives, including benchmark resources | Check sizes, entry counts, download URLs and SHA-256 hashes |
+| `modelangelo`, `cryoatom`, `e3_cryofold`, `emprot`, `af3` | Original per-method result tables | Access the released CSV columns and values |
 
-| Component | Entries | Data archives |
-|---|---:|---|
-| Training Part 1 | 4,164 | 17 independent archives |
-| Training Part 2 | 4,164 | 17 independent archives |
-| Training Part 3 | 4,165 | 17 independent archives |
-| Training Part 4 | 4,164 | 17 independent archives |
-| Test set | 1,152 | Two independent archives: 618 benchmark targets and 534 remaining test entries |
-| Benchmark resources | Results and predictions for the 618 targets, metadata, and 10 examples | Resource files described below |
+The catalog contains metadata and archive links; map, CIF and FASTA payloads reside in the archives. [Field definitions and provenance](docs/data_dictionary.md) explain the tables and previews.
 
-Training parts are disjoint download partitions of the same fixed training split. Each EMD entry is kept complete within one archive. The 10 examples duplicate entries from the main release and do not increase the dataset count.
+## Download the data you need
 
-## Training and test archives
+| Component | Entries | Archives | Compressed size | Files |
+|---|---:|---:|---:|---|
+| Training Part 1 | 4,164 | 17 | 560.0 GB | [Browse](https://huggingface.co/datasets/markfaa/CryoAtlas/tree/main/train/part01) |
+| Training Part 2 | 4,164 | 17 | 563.0 GB | [Browse](https://huggingface.co/datasets/markfaa/CryoAtlas/tree/main/train/part02) |
+| Training Part 3 | 4,165 | 17 | 564.9 GB | [Browse](https://huggingface.co/datasets/markfaa/CryoAtlas/tree/main/train/part03) |
+| Training Part 4 | 4,164 | 17 | 575.8 GB | [Browse](https://huggingface.co/datasets/markfaa/CryoAtlas/tree/main/train/part04) |
+| Benchmark test subset | 618 | 1 | 81.9 GB | [Download](https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/test/CryoAtlas_v1.0_test_benchmark_618.tar.gz) |
+| Remaining test entries | 534 | 1 | 69.0 GB | [Download](https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/test/CryoAtlas_v1.0_test_remaining_534.tar.gz) |
+| Benchmark resources | Results, predictions and 10 examples | 1 | 1.74 GB | [Download](https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/resources/CryoAtlas_v1.0_benchmark_resources.tar.gz) |
 
-Training archives are named:
+The four training parts are disjoint download partitions of one fixed split. **618 + 534 = 1,152 test entries**: the benchmark is a subset of the test set, not an additional split. The 618 targets support paired comparisons after training-overlap filtering and selection of targets with valid outputs from the evaluated learning-based methods. The remaining entries complete the fixed test split.
 
-```text
-CryoAtlas_v1.0_train_part01_shard01of17.tar.gz
-...
-CryoAtlas_v1.0_train_part04_shard17of17.tar.gz
-```
+For a first look, use the online previews or the **1.74 GB benchmark resource archive**. To locate a particular entry or download a selected training part, use the [download helper](scripts/download.py):
 
-The test set is provided as:
+```bash
+pip install -U huggingface_hub
+hf download markfaa/CryoAtlas scripts/download.py --repo-type dataset --local-dir .
 
-```text
-CryoAtlas_v1.0_test_benchmark_618.tar.gz
-CryoAtlas_v1.0_test_remaining_534.tar.gz
+# Show the archive and size for one entry; do not download the archive yet.
+python scripts/download.py --emd EMD-0719 --dry-run
+
+# Download Training Part 1 and verify each archive.
+python scripts/download.py --part 1 --destination CryoAtlas-downloads
 ```
 
-The 618-target subset supports paired cross-method comparison after training-overlap filtering and selection of targets with valid outputs from all evaluated learning-based methods. The remaining 534 entries complete the original test split; their exclusion from the common benchmark can reflect eligibility or completion criteria.
+`--emd` selects the **complete archive containing that entry**. More options, direct downloads, extraction instructions and checksum commands are in the [download guide](docs/downloads.md).
 
-Every archive can be extracted independently. Extract archives into the same parent directory to reconstruct:
+## Repository layout
 
 ```text
-CryoAtlas_v1.0/
-├── train/EMD-.../
-├── test/EMD-.../
-└── metadata/
+train/part01/ ... part04/    68 training archives
+test/                       Two test archives and a test-set README
+resources/                  Benchmark resource archive
+benchmark/                  Unified results and original per-method CSVs
+catalog/                    Searchable entry index, archive index and EMD-to-archive lookup
+previews/                   Lightweight density images and image-viewer table
+metadata/                   Source accessions, fixed split lists and per-part inventories
+checksums/                  Archive hashes; SHA256SUMS uses repository-relative paths
+docs/                       Download instructions and field definitions
+scripts/                    Selective download and checksum verification
+assets/                     Dataset overview and example gallery
 ```
 
-Each entry contains:
+## What is inside an entry?
+
+Training and test archives extract into `CryoAtlas_v1.0/train/EMD-.../` and `CryoAtlas_v1.0/test/EMD-.../`. Each archive is independently extractable; extract the desired archives into the same parent directory.
 
 | File | Contents |
 |---|---|
 | `emd_<id>_fullmap.map` | Experimental density map in MRC/CCP4 format |
-| `<pdb>.cif` | Corresponding atomic structure and deposited metadata |
+| `<pdb>.cif` | Atomic structure with deposited metadata |
 | `<pdb>.fasta` | Original protein sequence records |
-| `<pdb>_aligned.fasta` | Aligned sequence representation, including gap characters where applicable |
-| `<pdb>_trimmed.fasta` | Terminal-trimmed sequence representation |
-| `<pdb>_aligned.cif`, where provided | Structure aligned to the density-map frame |
+| `<pdb>_aligned.fasta` | Aligned sequences, including gap characters where applicable |
+| `<pdb>_trimmed.fasta` | Terminal-trimmed sequences |
+| `<pdb>_aligned.cif`, where supplied | Structure aligned to the density-map frame |
 
-Density maps are included in the training and test archives. Original scientific file contents are preserved. For map-frame comparisons, use the aligned CIF when supplied.
+Use the aligned CIF for map-frame comparisons when it is provided. The per-archive inventories record the exact files and hashes for each entry.
 
-## Benchmark resource files
+## Benchmark resources and examples
 
-```text
-resources/
-├── Benchmark_outcome/        Five CSVs, each containing 618 target-level rows
-├── Benchmark_rawdata/       Five method folders, each containing 618 predicted CIFs
-│   ├── AF3_final_cif/
-│   ├── CryoAtom_final_cif/
-│   ├── E3-CryoFold_final_cif/
-│   ├── EMprot_final_cif/
-│   └── ModelAngelo_final_cif/
-├── metadata/
-│   ├── Train.json
-│   ├── Test.json
-│   ├── Benchmark_618.json
-│   ├── Align_cif.json
-│   ├── source_accessions.csv
-│   └── resources.sha256
-├── sample/
-│   ├── Dataset/             10 self-contained example entries
-│   └── sample_selection.md
-└── README.md                This project-level README
-```
+The resource archive extracts into **`CryoAtlas_v1.0_resources/`**. It contains:
 
-The prediction folders and evaluation CSVs cover the same 618 EMD IDs listed in `metadata/Benchmark_618.json`. CryoAtom, E3-CryoFold, EMProt, and ModelAngelo are density-guided modeling methods; AF3 provides a sequence-based reference.
+- `Benchmark_outcome/`: the five original evaluation CSVs, with 618 target rows each.
+- `Benchmark_rawdata/`: predicted CIF files from AF3, CryoAtom, E3-CryoFold, EMProt and ModelAngelo, with 618 targets per method.
+- `metadata/`: source accessions, fixed split lists, alignment availability and resource-file checksums.
+- `sample/Dataset/`: 10 self-contained examples with maps, structures, sequences and available predictions.
 
-The CSVs include target information and released measurements of structural accuracy, density agreement, and stereochemical quality, together with the corresponding composite scores. Each CSV has 618 target rows; aggregate statistics can be calculated from these rows.
+The same five CSVs are available directly in this repository under `benchmark/results/`. The unified Viewer table adds a method column and removes method prefixes from column names; it preserves the released measurements. AF3 is a sequence-based reference; the other four methods use density maps.
 
-## Metadata and integrity checks
+<a id="examples"></a>
+### Example gallery
 
-The resource-level metadata provides:
+Nine examples belong to the 618-target benchmark. **EMD-0719 belongs to training**, includes an aligned CIF, and has no baseline predictions in the example package. These examples also occur in the full release and do not add new entries.
 
-- `Train.json`: the 16,657 training EMD IDs.
-- `Test.json`: the complete 1,152 test EMD IDs.
-- `Benchmark_618.json`: the 618 common benchmark EMD IDs.
-- `Align_cif.json`: entries with an additional aligned structure file.
-- `source_accessions.csv`: EMDB/PDB accessions, split membership, source map URLs, and descriptive metadata for all 17,809 entries.
+[Open the image Viewer](https://huggingface.co/datasets/markfaa/CryoAtlas/viewer/examples/examples)
 
-Each training part and test subset also supplies its own EMD list and file inventory. Archive-level `.sha256` files verify complete downloads; inventories provide per-file SHA256 hashes for the extracted scientific files. The inventories use paths relative to the archive extraction directory.
+<details>
+<summary>Show all 10 density previews</summary>
 
-For this resource package, run the following from the directory containing `Benchmark_outcome/`, `Benchmark_rawdata/`, and `metadata/`:
+![Maximum-intensity projections of the 10 released examples](assets/example_gallery.png)
 
-```bash
-sha256sum -c metadata/resources.sha256
-```
+</details>
+
+Previews show maximum-intensity projections along the three stored array axes. Each view uses its own 2nd–99.5th percentile display range. They are visual summaries, not aligned structure overlays or quantitative map-quality measurements. Original maps are available in the resource archive and full release.
 
-## Examples
+## Provenance, scope and license
 
-Start with `sample/Dataset/` to inspect a small set of density maps, structures, sequences, and available predictions. It contains nine benchmark entries and one training entry, `EMD-0719`, which illustrates a provided aligned structure. This training example has no baseline predictions in the sample. The examples do not define a separate evaluation split.
+[Source accessions](metadata/source_accessions.csv) identify EMDB/PDB records and original map URLs. [Fixed split lists](https://huggingface.co/datasets/markfaa/CryoAtlas/tree/main/metadata/splits), per-part EMD lists, file inventories and checksums support reproducible retrieval. The overview and catalog use the published metadata; no scientific data or benchmark measurements were modified during repository organization.
 
-## License
+CryoAtlas supports protein atomic model building from cryo-EM maps. The benchmark results describe the released methods and evaluation subset; the 10 examples are for inspection, not an independent evaluation set.
 
-The dataset and benchmark resources are distributed under [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/). Original source attribution and deposited metadata are retained in the files and accession table.
+The dataset and benchmark resources are distributed under [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/). Source attribution and deposited metadata are retained.

benchmark/results.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16abf04d09430ce018ef706ef66d8d64afd7d3bd73c45153fbdda9d514bc4899
+size 504373

catalog/archives.json

@@ -0,0 +1,641 @@
+[
+  {
+    "path": "resources/CryoAtlas_v1.0_benchmark_resources.tar.gz",
+    "bytes": 1737458410,
+    "sha256": "e11e9ebd96141520c10cdfa3d8c5e16036a586a413e4ac16f437385fdbbde29a",
+    "component": "benchmark_resources",
+    "entries": null,
+    "size_gb": 1.737,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/resources/CryoAtlas_v1.0_benchmark_resources.tar.gz"
+  },
+  {
+    "path": "test/CryoAtlas_v1.0_test_benchmark_618.tar.gz",
+    "bytes": 81944559072,
+    "sha256": "52e9c0b3a21f60ee2402b4168faae2f16bde4d91dc4568fb0ee8be63ec078137",
+    "component": "test_benchmark_618",
+    "entries": 618,
+    "size_gb": 81.945,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/test/CryoAtlas_v1.0_test_benchmark_618.tar.gz"
+  },
+  {
+    "path": "test/CryoAtlas_v1.0_test_remaining_534.tar.gz",
+    "bytes": 68984944750,
+    "sha256": "28eb24e148d426662f02cf657d07574211f5142f4c9394a1c442a83831b73601",
+    "component": "test_remaining_534",
+    "entries": 534,
+    "size_gb": 68.985,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/test/CryoAtlas_v1.0_test_remaining_534.tar.gz"
+  },
+  {
+    "path": "train/part01/CryoAtlas_v1.0_train_part01_shard01of17.tar.gz",
+    "bytes": 30371946668,
+    "sha256": "36a3a9d2e9f38fb96f0a3d8ad52b67a9a53d95fa1b1b6bd49487c750eb201dad",
+    "component": "train_part01",
+    "entries": 243,
+    "size_gb": 30.372,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part01/CryoAtlas_v1.0_train_part01_shard01of17.tar.gz"
+  },
+  {
+    "path": "train/part01/CryoAtlas_v1.0_train_part01_shard02of17.tar.gz",
+    "bytes": 34000397125,
+    "sha256": "6c6af95332efb217b38468268323cb62d236d2bd74d1a62723daa48e866da798",
+    "component": "train_part01",
+    "entries": 245,
+    "size_gb": 34.0,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part01/CryoAtlas_v1.0_train_part01_shard02of17.tar.gz"
+  },
+  {
+    "path": "train/part01/CryoAtlas_v1.0_train_part01_shard03of17.tar.gz",
+    "bytes": 33827895370,
+    "sha256": "5a18f48be8dbcaf66a86a5df9da826c233accbf239531fbedbf0abb1f871ce41",
+    "component": "train_part01",
+    "entries": 245,
+    "size_gb": 33.828,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part01/CryoAtlas_v1.0_train_part01_shard03of17.tar.gz"
+  },
+  {
+    "path": "train/part01/CryoAtlas_v1.0_train_part01_shard04of17.tar.gz",
+    "bytes": 33098284521,
+    "sha256": "967ddbdbc59419e717a454b8dd13f586efab7e15cd6d45d4b9916e9038bba462",
+    "component": "train_part01",
+    "entries": 245,
+    "size_gb": 33.098,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part01/CryoAtlas_v1.0_train_part01_shard04of17.tar.gz"
+  },
+  {
+    "path": "train/part01/CryoAtlas_v1.0_train_part01_shard05of17.tar.gz",
+    "bytes": 30400394779,
+    "sha256": "d092a07e3b1f138c9bfbd106234e766f22e6a1c1312c495b99726d0390b4da94",
+    "component": "train_part01",
+    "entries": 245,
+    "size_gb": 30.4,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part01/CryoAtlas_v1.0_train_part01_shard05of17.tar.gz"
+  },
+  {
+    "path": "train/part01/CryoAtlas_v1.0_train_part01_shard06of17.tar.gz",
+    "bytes": 35169290203,
+    "sha256": "15e1c8ed145f1b8563621933e19d1869bd4aa0c9201421d69573438833aa2c99",
+    "component": "train_part01",
+    "entries": 245,
+    "size_gb": 35.169,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part01/CryoAtlas_v1.0_train_part01_shard06of17.tar.gz"
+  },
+  {
+    "path": "train/part01/CryoAtlas_v1.0_train_part01_shard07of17.tar.gz",
+    "bytes": 30223818948,
+    "sha256": "37464711baeca7e734e7017d731287dda237e88de03a940215038cc0ec65c90d",
+    "component": "train_part01",
+    "entries": 245,
+    "size_gb": 30.224,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part01/CryoAtlas_v1.0_train_part01_shard07of17.tar.gz"
+  },
+  {
+    "path": "train/part01/CryoAtlas_v1.0_train_part01_shard08of17.tar.gz",
+    "bytes": 32053694620,
+    "sha256": "4a07e633b113165552aa5f0866b0828494b894e2638fc6141c2318286a8b946e",
+    "component": "train_part01",
+    "entries": 245,
+    "size_gb": 32.054,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part01/CryoAtlas_v1.0_train_part01_shard08of17.tar.gz"
+  },
+  {
+    "path": "train/part01/CryoAtlas_v1.0_train_part01_shard09of17.tar.gz",
+    "bytes": 33552941161,
+    "sha256": "994a73d22ea29aead14b93b34160f30b4ad3df489c574bda709ad4a3c67b8bbc",
+    "component": "train_part01",
+    "entries": 245,
+    "size_gb": 33.553,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part01/CryoAtlas_v1.0_train_part01_shard09of17.tar.gz"
+  },
+  {
+    "path": "train/part01/CryoAtlas_v1.0_train_part01_shard10of17.tar.gz",
+    "bytes": 34493122851,
+    "sha256": "9828105abf34c804ef5db5c8a15a9d54196fad4b0a2c2d4ff3b83cbfc444cb3c",
+    "component": "train_part01",
+    "entries": 245,
+    "size_gb": 34.493,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part01/CryoAtlas_v1.0_train_part01_shard10of17.tar.gz"
+  },
+  {
+    "path": "train/part01/CryoAtlas_v1.0_train_part01_shard11of17.tar.gz",
+    "bytes": 32850814922,
+    "sha256": "296067e7cec65f4e6a96c1a566eb03ef4cc153d7ec4af9bb32f8272796fafeb3",
+    "component": "train_part01",
+    "entries": 245,
+    "size_gb": 32.851,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part01/CryoAtlas_v1.0_train_part01_shard11of17.tar.gz"
+  },
+  {
+    "path": "train/part01/CryoAtlas_v1.0_train_part01_shard12of17.tar.gz",
+    "bytes": 32989073633,
+    "sha256": "de8bab4748c3ff5525e2d15b8cd67c8d62f2b4fdcbec9093093c254be69f12a9",
+    "component": "train_part01",
+    "entries": 245,
+    "size_gb": 32.989,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part01/CryoAtlas_v1.0_train_part01_shard12of17.tar.gz"
+  },
+  {
+    "path": "train/part01/CryoAtlas_v1.0_train_part01_shard13of17.tar.gz",
+    "bytes": 36291563240,
+    "sha256": "b50423a080c350dd98244f91d925ed4bb8b3503129814221b772286d5e30da8b",
+    "component": "train_part01",
+    "entries": 246,
+    "size_gb": 36.292,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part01/CryoAtlas_v1.0_train_part01_shard13of17.tar.gz"
+  },
+  {
+    "path": "train/part01/CryoAtlas_v1.0_train_part01_shard14of17.tar.gz",
+    "bytes": 31837546676,
+    "sha256": "798b6d6043278e3c54c58891c92cd691b3c7cf3aa5181025b0096fa5f620b98d",
+    "component": "train_part01",
+    "entries": 245,
+    "size_gb": 31.838,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part01/CryoAtlas_v1.0_train_part01_shard14of17.tar.gz"
+  },
+  {
+    "path": "train/part01/CryoAtlas_v1.0_train_part01_shard15of17.tar.gz",
+    "bytes": 33348487675,
+    "sha256": "4d7786b5aaa04ddc188b7272f47fc6ff73b23fc546f0876a901ed581902856e3",
+    "component": "train_part01",
+    "entries": 245,
+    "size_gb": 33.348,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part01/CryoAtlas_v1.0_train_part01_shard15of17.tar.gz"
+  },
+  {
+    "path": "train/part01/CryoAtlas_v1.0_train_part01_shard16of17.tar.gz",
+    "bytes": 33774555969,
+    "sha256": "be450f7ac2430d6f89e0c824ba908062d86056e26f1591a59363ca5c2b9f40d2",
+    "component": "train_part01",
+    "entries": 245,
+    "size_gb": 33.775,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part01/CryoAtlas_v1.0_train_part01_shard16of17.tar.gz"
+  },
+  {
+    "path": "train/part01/CryoAtlas_v1.0_train_part01_shard17of17.tar.gz",
+    "bytes": 31751316298,
+    "sha256": "6c3076d0499f060f739fd15f95a169183f1f56ab5a424ddb2005fdb2d374feee",
+    "component": "train_part01",
+    "entries": 245,
+    "size_gb": 31.751,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part01/CryoAtlas_v1.0_train_part01_shard17of17.tar.gz"
+  },
+  {
+    "path": "train/part02/CryoAtlas_v1.0_train_part02_shard01of17.tar.gz",
+    "bytes": 31016559239,
+    "sha256": "6da26eef647a88286802b35f65d2f0c472716630a8ce68bb37618c52b742c077",
+    "component": "train_part02",
+    "entries": 243,
+    "size_gb": 31.017,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part02/CryoAtlas_v1.0_train_part02_shard01of17.tar.gz"
+  },
+  {
+    "path": "train/part02/CryoAtlas_v1.0_train_part02_shard02of17.tar.gz",
+    "bytes": 32242759447,
+    "sha256": "6f45011cf5d7f5c6b70a6046def666f5ef7b1eeb01df235fee2d93d6656fb0b0",
+    "component": "train_part02",
+    "entries": 244,
+    "size_gb": 32.243,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part02/CryoAtlas_v1.0_train_part02_shard02of17.tar.gz"
+  },
+  {
+    "path": "train/part02/CryoAtlas_v1.0_train_part02_shard03of17.tar.gz",
+    "bytes": 33924646273,
+    "sha256": "64b4aa3ca44607d2b12430287d3f93c832a07f39d5fdaa0483875e49a497b194",
+    "component": "train_part02",
+    "entries": 245,
+    "size_gb": 33.925,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part02/CryoAtlas_v1.0_train_part02_shard03of17.tar.gz"
+  },
+  {
+    "path": "train/part02/CryoAtlas_v1.0_train_part02_shard04of17.tar.gz",
+    "bytes": 30201248636,
+    "sha256": "3011ada125047ea8f2b4a1ae1fe409ead0eeb89bf743709c6a70c452900436a1",
+    "component": "train_part02",
+    "entries": 245,
+    "size_gb": 30.201,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part02/CryoAtlas_v1.0_train_part02_shard04of17.tar.gz"
+  },
+  {
+    "path": "train/part02/CryoAtlas_v1.0_train_part02_shard05of17.tar.gz",
+    "bytes": 34147087227,
+    "sha256": "05ee21467213de4c88a16185a25522d1f8b971f5e9122498c838cee9148e0e02",
+    "component": "train_part02",
+    "entries": 245,
+    "size_gb": 34.147,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part02/CryoAtlas_v1.0_train_part02_shard05of17.tar.gz"
+  },
+  {
+    "path": "train/part02/CryoAtlas_v1.0_train_part02_shard06of17.tar.gz",
+    "bytes": 34924736968,
+    "sha256": "02239d96a6badb87b77b2bf02ff3016a0028adfeaa3ee5fe7de623483fb9b089",
+    "component": "train_part02",
+    "entries": 245,
+    "size_gb": 34.925,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part02/CryoAtlas_v1.0_train_part02_shard06of17.tar.gz"
+  },
+  {
+    "path": "train/part02/CryoAtlas_v1.0_train_part02_shard07of17.tar.gz",
+    "bytes": 34841268894,
+    "sha256": "6120cc53bd5c4d7c0dd2b6b73e0b2428dfda99632fb930140b76a3c8268fd0d0",
+    "component": "train_part02",
+    "entries": 245,
+    "size_gb": 34.841,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part02/CryoAtlas_v1.0_train_part02_shard07of17.tar.gz"
+  },
+  {
+    "path": "train/part02/CryoAtlas_v1.0_train_part02_shard08of17.tar.gz",
+    "bytes": 33742742870,
+    "sha256": "bbd1f59f8d0ad638121fc33aaf99508377855563c6da5f5e2374e956d6ddb799",
+    "component": "train_part02",
+    "entries": 245,
+    "size_gb": 33.743,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part02/CryoAtlas_v1.0_train_part02_shard08of17.tar.gz"
+  },
+  {
+    "path": "train/part02/CryoAtlas_v1.0_train_part02_shard09of17.tar.gz",
+    "bytes": 30124309298,
+    "sha256": "4e8e8ef91aa4482b5ecdfff909767d4bd60e7f46e98fe55920fdbda1fb05fbab",
+    "component": "train_part02",
+    "entries": 245,
+    "size_gb": 30.124,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part02/CryoAtlas_v1.0_train_part02_shard09of17.tar.gz"
+  },
+  {
+    "path": "train/part02/CryoAtlas_v1.0_train_part02_shard10of17.tar.gz",
+    "bytes": 33054254496,
+    "sha256": "a87d2e889ba9e8efc9552dbf4f082d22acb3212f87070836cd02c682ba9c4927",
+    "component": "train_part02",
+    "entries": 246,
+    "size_gb": 33.054,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part02/CryoAtlas_v1.0_train_part02_shard10of17.tar.gz"
+  },
+  {
+    "path": "train/part02/CryoAtlas_v1.0_train_part02_shard11of17.tar.gz",
+    "bytes": 32172956143,
+    "sha256": "5cebcfa3af6b12636c5f3ef8e1314e12520de279d0297774cea52828615986aa",
+    "component": "train_part02",
+    "entries": 245,
+    "size_gb": 32.173,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part02/CryoAtlas_v1.0_train_part02_shard11of17.tar.gz"
+  },
+  {
+    "path": "train/part02/CryoAtlas_v1.0_train_part02_shard12of17.tar.gz",
+    "bytes": 34462829811,
+    "sha256": "2cb2469a5b44723a631724957f401c34388b6824a9598cf24212e0b5d2ea85e7",
+    "component": "train_part02",
+    "entries": 246,
+    "size_gb": 34.463,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part02/CryoAtlas_v1.0_train_part02_shard12of17.tar.gz"
+  },
+  {
+    "path": "train/part02/CryoAtlas_v1.0_train_part02_shard13of17.tar.gz",
+    "bytes": 35527447415,
+    "sha256": "b8735ba506d5a8b4b75f31d9df9664348bbcf665b0df18c3a999cd602383acfc",
+    "component": "train_part02",
+    "entries": 245,
+    "size_gb": 35.527,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part02/CryoAtlas_v1.0_train_part02_shard13of17.tar.gz"
+  },
+  {
+    "path": "train/part02/CryoAtlas_v1.0_train_part02_shard14of17.tar.gz",
+    "bytes": 32954614871,
+    "sha256": "6c6a8c09016f78104a8ddbd6df18e5645aabd110ee903f2b17a58434505345b3",
+    "component": "train_part02",
+    "entries": 245,
+    "size_gb": 32.955,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part02/CryoAtlas_v1.0_train_part02_shard14of17.tar.gz"
+  },
+  {
+    "path": "train/part02/CryoAtlas_v1.0_train_part02_shard15of17.tar.gz",
+    "bytes": 32310008052,
+    "sha256": "bd5390114875cace524f5d581e06571bfef2e2768fe59410a49df9b77ca1b05d",
+    "component": "train_part02",
+    "entries": 245,
+    "size_gb": 32.31,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part02/CryoAtlas_v1.0_train_part02_shard15of17.tar.gz"
+  },
+  {
+    "path": "train/part02/CryoAtlas_v1.0_train_part02_shard16of17.tar.gz",
+    "bytes": 33115045320,
+    "sha256": "c2ea6c43784a12f49572c4bce631147cf93a9b89df0c7bfb4514910e011a7e47",
+    "component": "train_part02",
+    "entries": 245,
+    "size_gb": 33.115,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part02/CryoAtlas_v1.0_train_part02_shard16of17.tar.gz"
+  },
+  {
+    "path": "train/part02/CryoAtlas_v1.0_train_part02_shard17of17.tar.gz",
+    "bytes": 34253390980,
+    "sha256": "59f557ce9f059838c8cabcb22501c657cfe2564c472d6022eaa8a3440a4b62c2",
+    "component": "train_part02",
+    "entries": 245,
+    "size_gb": 34.253,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part02/CryoAtlas_v1.0_train_part02_shard17of17.tar.gz"
+  },
+  {
+    "path": "train/part03/CryoAtlas_v1.0_train_part03_shard01of17.tar.gz",
+    "bytes": 30233337854,
+    "sha256": "57401fef3a177e6d9407119e683e46db45e0ecea7b474459a8ce5409b905e6cf",
+    "component": "train_part03",
+    "entries": 243,
+    "size_gb": 30.233,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part03/CryoAtlas_v1.0_train_part03_shard01of17.tar.gz"
+  },
+  {
+    "path": "train/part03/CryoAtlas_v1.0_train_part03_shard02of17.tar.gz",
+    "bytes": 32820697427,
+    "sha256": "81479f9c0daef0725eb906ae1a650e9f499bd73bf388db20e9a6c7ca96723811",
+    "component": "train_part03",
+    "entries": 245,
+    "size_gb": 32.821,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part03/CryoAtlas_v1.0_train_part03_shard02of17.tar.gz"
+  },
+  {
+    "path": "train/part03/CryoAtlas_v1.0_train_part03_shard03of17.tar.gz",
+    "bytes": 31843081695,
+    "sha256": "f27580c285066bab04d123627d79d40d96ed378e2b6e34fcd31367a042f0c179",
+    "component": "train_part03",
+    "entries": 245,
+    "size_gb": 31.843,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part03/CryoAtlas_v1.0_train_part03_shard03of17.tar.gz"
+  },
+  {
+    "path": "train/part03/CryoAtlas_v1.0_train_part03_shard04of17.tar.gz",
+    "bytes": 33239372982,
+    "sha256": "675e2c25676aa17a0d595505351f88b94925869eeedfbd0641bd8766699e3e7b",
+    "component": "train_part03",
+    "entries": 245,
+    "size_gb": 33.239,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part03/CryoAtlas_v1.0_train_part03_shard04of17.tar.gz"
+  },
+  {
+    "path": "train/part03/CryoAtlas_v1.0_train_part03_shard05of17.tar.gz",
+    "bytes": 32815760485,
+    "sha256": "88187109ff8db99457b832d0c8f6943f4c39c53507b894dd64aecf8915d87a6f",
+    "component": "train_part03",
+    "entries": 245,
+    "size_gb": 32.816,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part03/CryoAtlas_v1.0_train_part03_shard05of17.tar.gz"
+  },
+  {
+    "path": "train/part03/CryoAtlas_v1.0_train_part03_shard06of17.tar.gz",
+    "bytes": 35665196591,
+    "sha256": "637a00c3ea90f67c807133c5cc4726f8f3106fa62832495e792027210fa5d105",
+    "component": "train_part03",
+    "entries": 245,
+    "size_gb": 35.665,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part03/CryoAtlas_v1.0_train_part03_shard06of17.tar.gz"
+  },
+  {
+    "path": "train/part03/CryoAtlas_v1.0_train_part03_shard07of17.tar.gz",
+    "bytes": 31436362333,
+    "sha256": "74c866bfa5cfb7df50c4ff39afaedd7c30e7ba4ab652eb73ba8094ee270046e6",
+    "component": "train_part03",
+    "entries": 245,
+    "size_gb": 31.436,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part03/CryoAtlas_v1.0_train_part03_shard07of17.tar.gz"
+  },
+  {
+    "path": "train/part03/CryoAtlas_v1.0_train_part03_shard08of17.tar.gz",
+    "bytes": 32777172032,
+    "sha256": "9e54dfe5706262f9a69d02ecb129891f304bc3118b48e9e35b818bc1b3ff426c",
+    "component": "train_part03",
+    "entries": 246,
+    "size_gb": 32.777,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part03/CryoAtlas_v1.0_train_part03_shard08of17.tar.gz"
+  },
+  {
+    "path": "train/part03/CryoAtlas_v1.0_train_part03_shard09of17.tar.gz",
+    "bytes": 31804993774,
+    "sha256": "73a33b7a0df0061a0f90ff39bfab17e2b5386c0519e3ea90282502a3e1de0a07",
+    "component": "train_part03",
+    "entries": 245,
+    "size_gb": 31.805,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part03/CryoAtlas_v1.0_train_part03_shard09of17.tar.gz"
+  },
+  {
+    "path": "train/part03/CryoAtlas_v1.0_train_part03_shard10of17.tar.gz",
+    "bytes": 34505775216,
+    "sha256": "e39cf41c5025024531acbffae1b675d043a95df2128a1957174b0ebc770c76fa",
+    "component": "train_part03",
+    "entries": 245,
+    "size_gb": 34.506,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part03/CryoAtlas_v1.0_train_part03_shard10of17.tar.gz"
+  },
+  {
+    "path": "train/part03/CryoAtlas_v1.0_train_part03_shard11of17.tar.gz",
+    "bytes": 32931007974,
+    "sha256": "e57f14ac681eb20f22e28943190dad613a07c8799c039ae8bd610f85373e4dcc",
+    "component": "train_part03",
+    "entries": 245,
+    "size_gb": 32.931,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part03/CryoAtlas_v1.0_train_part03_shard11of17.tar.gz"
+  },
+  {
+    "path": "train/part03/CryoAtlas_v1.0_train_part03_shard12of17.tar.gz",
+    "bytes": 35042266013,
+    "sha256": "a277fd01827b29bc84215fb0558b4573d739f96eafe4dc2821db74c475d0c408",
+    "component": "train_part03",
+    "entries": 245,
+    "size_gb": 35.042,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part03/CryoAtlas_v1.0_train_part03_shard12of17.tar.gz"
+  },
+  {
+    "path": "train/part03/CryoAtlas_v1.0_train_part03_shard13of17.tar.gz",
+    "bytes": 33710847865,
+    "sha256": "1f52807daea938aee6a11ed2894cd48fa41749535e677735fbe0aad113c576eb",
+    "component": "train_part03",
+    "entries": 246,
+    "size_gb": 33.711,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part03/CryoAtlas_v1.0_train_part03_shard13of17.tar.gz"
+  },
+  {
+    "path": "train/part03/CryoAtlas_v1.0_train_part03_shard14of17.tar.gz",
+    "bytes": 33229362806,
+    "sha256": "d95b73c9d9bece3955b16b8a8aedd304e0e742a601e9db2620c9ad5d0d57b02e",
+    "component": "train_part03",
+    "entries": 245,
+    "size_gb": 33.229,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part03/CryoAtlas_v1.0_train_part03_shard14of17.tar.gz"
+  },
+  {
+    "path": "train/part03/CryoAtlas_v1.0_train_part03_shard15of17.tar.gz",
+    "bytes": 33336668599,
+    "sha256": "2543c5abcabc360d6a7a9b4ac30a3c617ffed40e4b811ed19e19f3f73c749dd7",
+    "component": "train_part03",
+    "entries": 245,
+    "size_gb": 33.337,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part03/CryoAtlas_v1.0_train_part03_shard15of17.tar.gz"
+  },
+  {
+    "path": "train/part03/CryoAtlas_v1.0_train_part03_shard16of17.tar.gz",
+    "bytes": 35574108073,
+    "sha256": "a7c8d2116f0ff1d25bc59529a1bbd45520102b86e184eb44d8f8eeadf1f1ab0c",
+    "component": "train_part03",
+    "entries": 245,
+    "size_gb": 35.574,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part03/CryoAtlas_v1.0_train_part03_shard16of17.tar.gz"
+  },
+  {
+    "path": "train/part03/CryoAtlas_v1.0_train_part03_shard17of17.tar.gz",
+    "bytes": 33952315184,
+    "sha256": "f424ce618c09eac632f9bd68ce93a9a0d93eeb2425934cceac4be8cd0a00ccb8",
+    "component": "train_part03",
+    "entries": 245,
+    "size_gb": 33.952,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part03/CryoAtlas_v1.0_train_part03_shard17of17.tar.gz"
+  },
+  {
+    "path": "train/part04/CryoAtlas_v1.0_train_part04_shard01of17.tar.gz",
+    "bytes": 31678464281,
+    "sha256": "3aee1e3de1424c0d728fda2ee9fad1836c2412eadead3ec7736eccd42950c6ef",
+    "component": "train_part04",
+    "entries": 245,
+    "size_gb": 31.678,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part04/CryoAtlas_v1.0_train_part04_shard01of17.tar.gz"
+  },
+  {
+    "path": "train/part04/CryoAtlas_v1.0_train_part04_shard02of17.tar.gz",
+    "bytes": 33766030829,
+    "sha256": "a1ab96ab7dbbb412984a092ba62d51cc4879fd63dfe0814dffcf8f0167e78dd4",
+    "component": "train_part04",
+    "entries": 244,
+    "size_gb": 33.766,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part04/CryoAtlas_v1.0_train_part04_shard02of17.tar.gz"
+  },
+  {
+    "path": "train/part04/CryoAtlas_v1.0_train_part04_shard03of17.tar.gz",
+    "bytes": 34509008925,
+    "sha256": "50c281b6ea706c30d9a1f281b296b4e4799a275de7f72a6b144eeea5c7966687",
+    "component": "train_part04",
+    "entries": 245,
+    "size_gb": 34.509,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part04/CryoAtlas_v1.0_train_part04_shard03of17.tar.gz"
+  },
+  {
+    "path": "train/part04/CryoAtlas_v1.0_train_part04_shard04of17.tar.gz",
+    "bytes": 33665064759,
+    "sha256": "aaf8e721e0073e79528e9154df0d3a4a544608953f23c573bfebcd865a85a416",
+    "component": "train_part04",
+    "entries": 245,
+    "size_gb": 33.665,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part04/CryoAtlas_v1.0_train_part04_shard04of17.tar.gz"
+  },
+  {
+    "path": "train/part04/CryoAtlas_v1.0_train_part04_shard05of17.tar.gz",
+    "bytes": 34872641168,
+    "sha256": "e2dd4d190e0e0f66f93ab8594240bc338fdd5babe4c41966413d67fa0bff4ace",
+    "component": "train_part04",
+    "entries": 245,
+    "size_gb": 34.873,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part04/CryoAtlas_v1.0_train_part04_shard05of17.tar.gz"
+  },
+  {
+    "path": "train/part04/CryoAtlas_v1.0_train_part04_shard06of17.tar.gz",
+    "bytes": 35047215019,
+    "sha256": "83f3aebc3b2bb2097c85b3f57ec2528a95eeef44e23f727eb2eebfca0cd07c3e",
+    "component": "train_part04",
+    "entries": 245,
+    "size_gb": 35.047,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part04/CryoAtlas_v1.0_train_part04_shard06of17.tar.gz"
+  },
+  {
+    "path": "train/part04/CryoAtlas_v1.0_train_part04_shard07of17.tar.gz",
+    "bytes": 32359813968,
+    "sha256": "b1d9921bb675f87082d6056555a0fd29d1c15e4157990de94865c190eb306444",
+    "component": "train_part04",
+    "entries": 245,
+    "size_gb": 32.36,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part04/CryoAtlas_v1.0_train_part04_shard07of17.tar.gz"
+  },
+  {
+    "path": "train/part04/CryoAtlas_v1.0_train_part04_shard08of17.tar.gz",
+    "bytes": 34430107319,
+    "sha256": "c87298e9753fc399fcc18f6685f10661e1d30ed17b126ed22b7c6a2d3ee77317",
+    "component": "train_part04",
+    "entries": 245,
+    "size_gb": 34.43,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part04/CryoAtlas_v1.0_train_part04_shard08of17.tar.gz"
+  },
+  {
+    "path": "train/part04/CryoAtlas_v1.0_train_part04_shard09of17.tar.gz",
+    "bytes": 33787658151,
+    "sha256": "ed2b9e36fd35791cfbc9aed5136de54c85d895d18a984204e3178d79bdbd2f05",
+    "component": "train_part04",
+    "entries": 245,
+    "size_gb": 33.788,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part04/CryoAtlas_v1.0_train_part04_shard09of17.tar.gz"
+  },
+  {
+    "path": "train/part04/CryoAtlas_v1.0_train_part04_shard10of17.tar.gz",
+    "bytes": 33006665794,
+    "sha256": "20d7363890059d6c816320b6baf1357cbef0628f8ebc6caae3862928c37c11e6",
+    "component": "train_part04",
+    "entries": 245,
+    "size_gb": 33.007,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part04/CryoAtlas_v1.0_train_part04_shard10of17.tar.gz"
+  },
+  {
+    "path": "train/part04/CryoAtlas_v1.0_train_part04_shard11of17.tar.gz",
+    "bytes": 33552210698,
+    "sha256": "f24d4d76fb338fa97fefb3224856c4b6c687a3e73b658e598c4e5d8a31b4ee04",
+    "component": "train_part04",
+    "entries": 245,
+    "size_gb": 33.552,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part04/CryoAtlas_v1.0_train_part04_shard11of17.tar.gz"
+  },
+  {
+    "path": "train/part04/CryoAtlas_v1.0_train_part04_shard12of17.tar.gz",
+    "bytes": 34651155420,
+    "sha256": "9eb8b99e89a80e8be141e9850abd54896929a13eaaa5a186436a13597ee5dd87",
+    "component": "train_part04",
+    "entries": 245,
+    "size_gb": 34.651,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part04/CryoAtlas_v1.0_train_part04_shard12of17.tar.gz"
+  },
+  {
+    "path": "train/part04/CryoAtlas_v1.0_train_part04_shard13of17.tar.gz",
+    "bytes": 33291056618,
+    "sha256": "aa9f412c09516b2ae928c7a319e3851081a389c022b6813be4e9eb23f86c3a47",
+    "component": "train_part04",
+    "entries": 245,
+    "size_gb": 33.291,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part04/CryoAtlas_v1.0_train_part04_shard13of17.tar.gz"
+  },
+  {
+    "path": "train/part04/CryoAtlas_v1.0_train_part04_shard14of17.tar.gz",
+    "bytes": 33976612450,
+    "sha256": "68499a57c8d37a0d459e655d96c101f4574ca0f187caddeea7dd21e60439374a",
+    "component": "train_part04",
+    "entries": 245,
+    "size_gb": 33.977,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part04/CryoAtlas_v1.0_train_part04_shard14of17.tar.gz"
+  },
+  {
+    "path": "train/part04/CryoAtlas_v1.0_train_part04_shard15of17.tar.gz",
+    "bytes": 32719692059,
+    "sha256": "1d9d1fd320fa18c685807eea2e972c65b3b231c7a1d3f31037d57d7f12b895a5",
+    "component": "train_part04",
+    "entries": 245,
+    "size_gb": 32.72,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part04/CryoAtlas_v1.0_train_part04_shard15of17.tar.gz"
+  },
+  {
+    "path": "train/part04/CryoAtlas_v1.0_train_part04_shard16of17.tar.gz",
+    "bytes": 34965649753,
+    "sha256": "d5cdbc690c0808c9b6c29dd26b305a66ad92b47f78cb6ad01295668a5a17aac8",
+    "component": "train_part04",
+    "entries": 245,
+    "size_gb": 34.966,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part04/CryoAtlas_v1.0_train_part04_shard16of17.tar.gz"
+  },
+  {
+    "path": "train/part04/CryoAtlas_v1.0_train_part04_shard17of17.tar.gz",
+    "bytes": 35529634479,
+    "sha256": "80387f48ab068793c8b4e446fa66aba7d8676a92f4baf843ea3199fa8bc05a1a",
+    "component": "train_part04",
+    "entries": 245,
+    "size_gb": 35.53,
+    "download_url": "https://huggingface.co/datasets/markfaa/CryoAtlas/resolve/main/train/part04/CryoAtlas_v1.0_train_part04_shard17of17.tar.gz"
+  }
+]

catalog/archives.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e27d3ea2ef8c8f3bf401e20cdb282b3aef58086c5b10eb2309976e6d606beac
+size 7740

catalog/benchmark_targets.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1cf3823f275584a7547bfa5df4bfafadf1313aee651a8ec1699a4a797673a1f
+size 49750

catalog/test.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f388fb1fa76f0d9b4ea4a5e6e6b1692088874a3f5244fd95b5c182a0ce65cf7
+size 87096

catalog/train.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc83fa8a5bcb86eecf22f22ea4908b0df8318b2974045bfe319ea765e1441a1e
+size 994252

checksums/SHA256SUMS

@@ -0,0 +1,71 @@
+e11e9ebd96141520c10cdfa3d8c5e16036a586a413e4ac16f437385fdbbde29a  resources/CryoAtlas_v1.0_benchmark_resources.tar.gz
+52e9c0b3a21f60ee2402b4168faae2f16bde4d91dc4568fb0ee8be63ec078137  test/CryoAtlas_v1.0_test_benchmark_618.tar.gz
+28eb24e148d426662f02cf657d07574211f5142f4c9394a1c442a83831b73601  test/CryoAtlas_v1.0_test_remaining_534.tar.gz
+36a3a9d2e9f38fb96f0a3d8ad52b67a9a53d95fa1b1b6bd49487c750eb201dad  train/part01/CryoAtlas_v1.0_train_part01_shard01of17.tar.gz
+6c6af95332efb217b38468268323cb62d236d2bd74d1a62723daa48e866da798  train/part01/CryoAtlas_v1.0_train_part01_shard02of17.tar.gz
+5a18f48be8dbcaf66a86a5df9da826c233accbf239531fbedbf0abb1f871ce41  train/part01/CryoAtlas_v1.0_train_part01_shard03of17.tar.gz
+967ddbdbc59419e717a454b8dd13f586efab7e15cd6d45d4b9916e9038bba462  train/part01/CryoAtlas_v1.0_train_part01_shard04of17.tar.gz
+d092a07e3b1f138c9bfbd106234e766f22e6a1c1312c495b99726d0390b4da94  train/part01/CryoAtlas_v1.0_train_part01_shard05of17.tar.gz
+15e1c8ed145f1b8563621933e19d1869bd4aa0c9201421d69573438833aa2c99  train/part01/CryoAtlas_v1.0_train_part01_shard06of17.tar.gz
+37464711baeca7e734e7017d731287dda237e88de03a940215038cc0ec65c90d  train/part01/CryoAtlas_v1.0_train_part01_shard07of17.tar.gz
+4a07e633b113165552aa5f0866b0828494b894e2638fc6141c2318286a8b946e  train/part01/CryoAtlas_v1.0_train_part01_shard08of17.tar.gz
+994a73d22ea29aead14b93b34160f30b4ad3df489c574bda709ad4a3c67b8bbc  train/part01/CryoAtlas_v1.0_train_part01_shard09of17.tar.gz
+9828105abf34c804ef5db5c8a15a9d54196fad4b0a2c2d4ff3b83cbfc444cb3c  train/part01/CryoAtlas_v1.0_train_part01_shard10of17.tar.gz
+296067e7cec65f4e6a96c1a566eb03ef4cc153d7ec4af9bb32f8272796fafeb3  train/part01/CryoAtlas_v1.0_train_part01_shard11of17.tar.gz
+de8bab4748c3ff5525e2d15b8cd67c8d62f2b4fdcbec9093093c254be69f12a9  train/part01/CryoAtlas_v1.0_train_part01_shard12of17.tar.gz
+b50423a080c350dd98244f91d925ed4bb8b3503129814221b772286d5e30da8b  train/part01/CryoAtlas_v1.0_train_part01_shard13of17.tar.gz
+798b6d6043278e3c54c58891c92cd691b3c7cf3aa5181025b0096fa5f620b98d  train/part01/CryoAtlas_v1.0_train_part01_shard14of17.tar.gz
+4d7786b5aaa04ddc188b7272f47fc6ff73b23fc546f0876a901ed581902856e3  train/part01/CryoAtlas_v1.0_train_part01_shard15of17.tar.gz
+be450f7ac2430d6f89e0c824ba908062d86056e26f1591a59363ca5c2b9f40d2  train/part01/CryoAtlas_v1.0_train_part01_shard16of17.tar.gz
+6c3076d0499f060f739fd15f95a169183f1f56ab5a424ddb2005fdb2d374feee  train/part01/CryoAtlas_v1.0_train_part01_shard17of17.tar.gz
+6da26eef647a88286802b35f65d2f0c472716630a8ce68bb37618c52b742c077  train/part02/CryoAtlas_v1.0_train_part02_shard01of17.tar.gz
+6f45011cf5d7f5c6b70a6046def666f5ef7b1eeb01df235fee2d93d6656fb0b0  train/part02/CryoAtlas_v1.0_train_part02_shard02of17.tar.gz
+64b4aa3ca44607d2b12430287d3f93c832a07f39d5fdaa0483875e49a497b194  train/part02/CryoAtlas_v1.0_train_part02_shard03of17.tar.gz
+3011ada125047ea8f2b4a1ae1fe409ead0eeb89bf743709c6a70c452900436a1  train/part02/CryoAtlas_v1.0_train_part02_shard04of17.tar.gz
+05ee21467213de4c88a16185a25522d1f8b971f5e9122498c838cee9148e0e02  train/part02/CryoAtlas_v1.0_train_part02_shard05of17.tar.gz
+02239d96a6badb87b77b2bf02ff3016a0028adfeaa3ee5fe7de623483fb9b089  train/part02/CryoAtlas_v1.0_train_part02_shard06of17.tar.gz
+6120cc53bd5c4d7c0dd2b6b73e0b2428dfda99632fb930140b76a3c8268fd0d0  train/part02/CryoAtlas_v1.0_train_part02_shard07of17.tar.gz
+bbd1f59f8d0ad638121fc33aaf99508377855563c6da5f5e2374e956d6ddb799  train/part02/CryoAtlas_v1.0_train_part02_shard08of17.tar.gz
+4e8e8ef91aa4482b5ecdfff909767d4bd60e7f46e98fe55920fdbda1fb05fbab  train/part02/CryoAtlas_v1.0_train_part02_shard09of17.tar.gz
+a87d2e889ba9e8efc9552dbf4f082d22acb3212f87070836cd02c682ba9c4927  train/part02/CryoAtlas_v1.0_train_part02_shard10of17.tar.gz
+5cebcfa3af6b12636c5f3ef8e1314e12520de279d0297774cea52828615986aa  train/part02/CryoAtlas_v1.0_train_part02_shard11of17.tar.gz
+2cb2469a5b44723a631724957f401c34388b6824a9598cf24212e0b5d2ea85e7  train/part02/CryoAtlas_v1.0_train_part02_shard12of17.tar.gz
+b8735ba506d5a8b4b75f31d9df9664348bbcf665b0df18c3a999cd602383acfc  train/part02/CryoAtlas_v1.0_train_part02_shard13of17.tar.gz
+6c6a8c09016f78104a8ddbd6df18e5645aabd110ee903f2b17a58434505345b3  train/part02/CryoAtlas_v1.0_train_part02_shard14of17.tar.gz
+bd5390114875cace524f5d581e06571bfef2e2768fe59410a49df9b77ca1b05d  train/part02/CryoAtlas_v1.0_train_part02_shard15of17.tar.gz
+c2ea6c43784a12f49572c4bce631147cf93a9b89df0c7bfb4514910e011a7e47  train/part02/CryoAtlas_v1.0_train_part02_shard16of17.tar.gz
+59f557ce9f059838c8cabcb22501c657cfe2564c472d6022eaa8a3440a4b62c2  train/part02/CryoAtlas_v1.0_train_part02_shard17of17.tar.gz
+57401fef3a177e6d9407119e683e46db45e0ecea7b474459a8ce5409b905e6cf  train/part03/CryoAtlas_v1.0_train_part03_shard01of17.tar.gz
+81479f9c0daef0725eb906ae1a650e9f499bd73bf388db20e9a6c7ca96723811  train/part03/CryoAtlas_v1.0_train_part03_shard02of17.tar.gz
+f27580c285066bab04d123627d79d40d96ed378e2b6e34fcd31367a042f0c179  train/part03/CryoAtlas_v1.0_train_part03_shard03of17.tar.gz
+675e2c25676aa17a0d595505351f88b94925869eeedfbd0641bd8766699e3e7b  train/part03/CryoAtlas_v1.0_train_part03_shard04of17.tar.gz
+88187109ff8db99457b832d0c8f6943f4c39c53507b894dd64aecf8915d87a6f  train/part03/CryoAtlas_v1.0_train_part03_shard05of17.tar.gz
+637a00c3ea90f67c807133c5cc4726f8f3106fa62832495e792027210fa5d105  train/part03/CryoAtlas_v1.0_train_part03_shard06of17.tar.gz
+74c866bfa5cfb7df50c4ff39afaedd7c30e7ba4ab652eb73ba8094ee270046e6  train/part03/CryoAtlas_v1.0_train_part03_shard07of17.tar.gz
+9e54dfe5706262f9a69d02ecb129891f304bc3118b48e9e35b818bc1b3ff426c  train/part03/CryoAtlas_v1.0_train_part03_shard08of17.tar.gz
+73a33b7a0df0061a0f90ff39bfab17e2b5386c0519e3ea90282502a3e1de0a07  train/part03/CryoAtlas_v1.0_train_part03_shard09of17.tar.gz
+e39cf41c5025024531acbffae1b675d043a95df2128a1957174b0ebc770c76fa  train/part03/CryoAtlas_v1.0_train_part03_shard10of17.tar.gz
+e57f14ac681eb20f22e28943190dad613a07c8799c039ae8bd610f85373e4dcc  train/part03/CryoAtlas_v1.0_train_part03_shard11of17.tar.gz
+a277fd01827b29bc84215fb0558b4573d739f96eafe4dc2821db74c475d0c408  train/part03/CryoAtlas_v1.0_train_part03_shard12of17.tar.gz
+1f52807daea938aee6a11ed2894cd48fa41749535e677735fbe0aad113c576eb  train/part03/CryoAtlas_v1.0_train_part03_shard13of17.tar.gz
+d95b73c9d9bece3955b16b8a8aedd304e0e742a601e9db2620c9ad5d0d57b02e  train/part03/CryoAtlas_v1.0_train_part03_shard14of17.tar.gz
+2543c5abcabc360d6a7a9b4ac30a3c617ffed40e4b811ed19e19f3f73c749dd7  train/part03/CryoAtlas_v1.0_train_part03_shard15of17.tar.gz
+a7c8d2116f0ff1d25bc59529a1bbd45520102b86e184eb44d8f8eeadf1f1ab0c  train/part03/CryoAtlas_v1.0_train_part03_shard16of17.tar.gz
+f424ce618c09eac632f9bd68ce93a9a0d93eeb2425934cceac4be8cd0a00ccb8  train/part03/CryoAtlas_v1.0_train_part03_shard17of17.tar.gz
+3aee1e3de1424c0d728fda2ee9fad1836c2412eadead3ec7736eccd42950c6ef  train/part04/CryoAtlas_v1.0_train_part04_shard01of17.tar.gz
+a1ab96ab7dbbb412984a092ba62d51cc4879fd63dfe0814dffcf8f0167e78dd4  train/part04/CryoAtlas_v1.0_train_part04_shard02of17.tar.gz
+50c281b6ea706c30d9a1f281b296b4e4799a275de7f72a6b144eeea5c7966687  train/part04/CryoAtlas_v1.0_train_part04_shard03of17.tar.gz
+aaf8e721e0073e79528e9154df0d3a4a544608953f23c573bfebcd865a85a416  train/part04/CryoAtlas_v1.0_train_part04_shard04of17.tar.gz
+e2dd4d190e0e0f66f93ab8594240bc338fdd5babe4c41966413d67fa0bff4ace  train/part04/CryoAtlas_v1.0_train_part04_shard05of17.tar.gz
+83f3aebc3b2bb2097c85b3f57ec2528a95eeef44e23f727eb2eebfca0cd07c3e  train/part04/CryoAtlas_v1.0_train_part04_shard06of17.tar.gz
+b1d9921bb675f87082d6056555a0fd29d1c15e4157990de94865c190eb306444  train/part04/CryoAtlas_v1.0_train_part04_shard07of17.tar.gz
+c87298e9753fc399fcc18f6685f10661e1d30ed17b126ed22b7c6a2d3ee77317  train/part04/CryoAtlas_v1.0_train_part04_shard08of17.tar.gz
+ed2b9e36fd35791cfbc9aed5136de54c85d895d18a984204e3178d79bdbd2f05  train/part04/CryoAtlas_v1.0_train_part04_shard09of17.tar.gz
+20d7363890059d6c816320b6baf1357cbef0628f8ebc6caae3862928c37c11e6  train/part04/CryoAtlas_v1.0_train_part04_shard10of17.tar.gz
+f24d4d76fb338fa97fefb3224856c4b6c687a3e73b658e598c4e5d8a31b4ee04  train/part04/CryoAtlas_v1.0_train_part04_shard11of17.tar.gz
+9eb8b99e89a80e8be141e9850abd54896929a13eaaa5a186436a13597ee5dd87  train/part04/CryoAtlas_v1.0_train_part04_shard12of17.tar.gz
+aa9f412c09516b2ae928c7a319e3851081a389c022b6813be4e9eb23f86c3a47  train/part04/CryoAtlas_v1.0_train_part04_shard13of17.tar.gz
+68499a57c8d37a0d459e655d96c101f4574ca0f187caddeea7dd21e60439374a  train/part04/CryoAtlas_v1.0_train_part04_shard14of17.tar.gz
+1d9d1fd320fa18c685807eea2e972c65b3b231c7a1d3f31037d57d7f12b895a5  train/part04/CryoAtlas_v1.0_train_part04_shard15of17.tar.gz
+d5cdbc690c0808c9b6c29dd26b305a66ad92b47f78cb6ad01295668a5a17aac8  train/part04/CryoAtlas_v1.0_train_part04_shard16of17.tar.gz
+80387f48ab068793c8b4e446fa66aba7d8676a92f4baf843ea3199fa8bc05a1a  train/part04/CryoAtlas_v1.0_train_part04_shard17of17.tar.gz

docs/data_dictionary.md

@@ -0,0 +1,47 @@
+# Tables and preview provenance
+
+## Entry catalog
+
+`catalog/train.parquet` and `catalog/test.parquet` contain one row per EMD entry. `catalog/benchmark_targets.parquet` is the 618-row subset of the test catalog.
+
+| Field | Meaning |
+|---|---|
+| `emd_id`, `pdb_id` | Released EMDB and corresponding PDB identifiers |
+| `split` | Fixed `train` or `test` membership |
+| `benchmark_618` | Whether the entry belongs to the common 618-target benchmark |
+| `resolution_angstrom` | Reported map resolution from `metadata/source_accessions.csv`, in Å |
+| `title` | Deposited description from the source-accession table |
+| `release_component` | Training download part or one of the two test archives |
+| `archive_path`, `download_url` | Repository-relative path and direct URL of the complete containing archive |
+| `entry_directory` | Entry directory inside the extracted training/test release |
+| `has_aligned_cif` | Availability recorded by the released `metadata/Align_cif.json` list |
+| `emdb_url`, `pdb_url`, `source_map_url` | Public accession pages and original map location |
+| `map_release_date`, `structure_determination_method`, `sample_type` | Fields preserved from the source-accession table |
+
+Membership is checked against the fixed split lists. Training archive assignments come from the frozen release plan; test assignments are checked against both test-subset EMD lists. Catalog rows contain no server paths, credentials, or upload logs.
+
+## Benchmark result table
+
+`benchmark/results.parquet` has one row per **EMD ID × method**: 618 targets and five methods, totaling 3,090 rows. It combines the five original CSVs under `benchmark/results/`. Original CSV values and bytes are preserved.
+
+- `method` identifies ModelAngelo, CryoAtom, E3-CryoFold, EMProt or AF3.
+- `emd_id`, `pdb_id`, `category`, `total_residues`, `resolution` and `n_chains` retain the released target descriptors.
+- Metric names lose only their method prefix: for example, `ModelAngelo_gt_score` becomes `gt_score` in a row whose `method` is `ModelAngelo`.
+- `tm_score`, `ca_rmsd`, residue coverage, sequence metrics and `gt_score` concern structure-to-reference agreement.
+- `cc_*`, `qscore`, `recall_qscore` and `map_score` describe the released map-agreement measurements.
+- Ramachandran, rotamer, Cβ, clash, bond/angle, MolProbity and `geom_*` columns contain the released stereochemical measurements and composite components.
+- `train_pident` and `train_bin` retain the identity annotations shipped in each method CSV. AF3 has no such columns in its source CSV, so the unified table leaves them null. Nulls are not zero measurements.
+
+No scores are recomputed or rescaled. Raw metrics use different units and directions; the table is a convenient view of the released evaluation outputs, not a new scoring protocol. When reproducing a published aggregate, use the paper's specified subset and aggregation rules.
+
+## Archive index and checksums
+
+`catalog/archives.json` and `catalog/archives.parquet` list the same 71 files. `bytes` is the exact compressed size; `size_gb` uses decimal GB (10⁹ bytes). `sha256` hashes the complete compressed file. `entries` counts the map–model–sequence entries in a training/test archive; it is null for the mixed benchmark resource package. `checksums/SHA256SUMS` gives the same hashes with repository-relative paths.
+
+## Density previews
+
+`previews/examples.parquet` contains an HF Image feature with embedded PNG bytes, plus example metadata and archive links. The ten maps are the existing examples in the benchmark resource package: nine benchmark-test entries and the training entry EMD-0719.
+
+Each image displays maximum-intensity projections along array axes 0, 1 and 2 of the stored MRC volume. Each projection is independently displayed between its 2nd and 99.5th percentiles, using a common color map. The axes refer to array storage order, not a standardized physical orientation. These display transformations do not alter the source maps and do not establish map–model alignment or quantify map quality.
+
+The overview figure counts entries in the fixed split lists and displays normalized histograms of reported resolutions from the source-accession table. All 17,809 resolutions are present. The plots and previews are derived browsing aids; the scientific archive files are unchanged.

docs/downloads.md

@@ -0,0 +1,82 @@
+# Download and verify CryoAtlas v1.0
+
+## Choose a component
+
+Use the [entry catalog](https://huggingface.co/datasets/markfaa/CryoAtlas/viewer/catalog/train) to find an EMD ID and its `archive_path`. [archives.json](../catalog/archives.json) lists all 71 downloadable archives, sizes and hashes. [entry_archives.json](../catalog/entry_archives.json) maps every one of the 17,809 EMD IDs to its archive.
+
+Archive sizes are approximately 30–36 GB for training shards, 81.9 GB for the 618-target test archive, 69.0 GB for the remaining test archive, and 1.74 GB for benchmark resources. The full training and test release is 2.415 TB compressed. Allow additional disk space for extraction.
+
+## Selective download helper
+
+```bash
+pip install -U huggingface_hub
+hf download markfaa/CryoAtlas scripts/download.py --repo-type dataset --local-dir .
+python scripts/download.py --part 1 --dry-run
+python scripts/download.py --part 1 2 --destination CryoAtlas-downloads
+python scripts/download.py --test benchmark --destination CryoAtlas-downloads
+python scripts/download.py --test remaining --destination CryoAtlas-downloads
+python scripts/download.py --resources --destination CryoAtlas-downloads
+python scripts/download.py --emd EMD-0719 --dry-run
+```
+
+Only the selected archives are downloaded. The helper pins a repository revision, checks each archive's byte count and computes SHA-256 after downloading. `--emd` retrieves an entire archive containing the entry; it does not extract or download just that entry. `--dry-run` downloads only the small indexes needed to show the selection. No login is required for this public repository.
+
+## Direct Python download
+
+```python
+from huggingface_hub import hf_hub_download
+
+path = hf_hub_download(
+    repo_id="markfaa/CryoAtlas",
+    repo_type="dataset",
+    filename="resources/CryoAtlas_v1.0_benchmark_resources.tar.gz",
+    local_dir="CryoAtlas-downloads",
+)
+```
+
+## Extract independently
+
+From the download destination:
+
+```bash
+mkdir -p extracted
+tar -xzf resources/CryoAtlas_v1.0_benchmark_resources.tar.gz -C extracted
+# Replace this path with the training or test archive you downloaded.
+tar -xzf train/part01/CryoAtlas_v1.0_train_part01_shard01of17.tar.gz -C extracted
+```
+
+Training/test archives reconstruct `extracted/CryoAtlas_v1.0/train/` and `extracted/CryoAtlas_v1.0/test/`. The resource archive reconstructs `extracted/CryoAtlas_v1.0_resources/`. Folder names inside the resource archive retain their original v1.0 spelling, including `Benchmark_outcome/` and `Benchmark_rawdata/`.
+
+Each training part also has a compact metadata archive under `metadata/train/partXX/`, with its EMD IDs and inventories. Individual `*_files.json` files in that directory describe each data archive. Test inventories are under `metadata/test/test_benchmark_618/` and `metadata/test/test_remaining_534/`.
+
+## Check integrity
+
+The helper verifies hashes automatically. To verify a manually downloaded archive, download [checksums/SHA256SUMS](../checksums/SHA256SUMS) and run from the repository-style download root:
+
+```bash
+# Linux: verify one downloaded archive.
+grep 'train/part01/CryoAtlas_v1.0_train_part01_shard01of17.tar.gz$' checksums/SHA256SUMS | sha256sum -c -
+# macOS: use shasum instead.
+grep 'train/part01/CryoAtlas_v1.0_train_part01_shard01of17.tar.gz$' checksums/SHA256SUMS | shasum -a 256 -c -
+```
+
+For a full download, use `sha256sum -c checksums/SHA256SUMS` (macOS: `shasum -a 256 -c checksums/SHA256SUMS`). The consolidated file covers 70 training/test archives and the resource archive; an intentionally partial download will otherwise report missing files.
+
+The older individual `.sha256` sidecars name the archive by basename. Run those from the directory containing that archive. The resource-file inventory `metadata/resources.sha256` applies **inside the extracted `CryoAtlas_v1.0_resources/` directory**, not to repository paths:
+
+```bash
+cd extracted/CryoAtlas_v1.0_resources
+sha256sum -c metadata/resources.sha256
+```
+
+## Load only the lightweight tables
+
+```python
+from datasets import load_dataset
+
+entries = load_dataset("markfaa/CryoAtlas", "catalog", split="test")
+results = load_dataset("markfaa/CryoAtlas", "benchmark_results", split="test")
+examples = load_dataset("markfaa/CryoAtlas", "examples", split="examples")
+```
+
+These configurations load indexes, released evaluation rows, or preview images. They do not download the multi-terabyte map archives. Original method configurations remain available, for example `load_dataset("markfaa/CryoAtlas", "modelangelo", split="test")`.

previews/examples.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6566f4481febc3c8bfbf8110a23425f539ef78e1445d4d71424f2aa0c086d05
+size 2014862

scripts/download.py

@@ -0,0 +1,72 @@
+#!/usr/bin/env python3
+"""Select CryoAtlas release archives, download them, and verify SHA-256."""
+import argparse
+import hashlib
+import json
+from pathlib import Path
+from huggingface_hub import HfApi, hf_hub_download
+
+REPO = 'markfaa/CryoAtlas'
+
+def select_archives(manifest, args, entry_map=None):
+    if args.part:
+        parts = {f'train_part{n:02d}' for n in args.part}
+        return [r for r in manifest if r['component'] in parts]
+    if args.test:
+        components = {'test_benchmark_618', 'test_remaining_534'} if args.test == 'all' else {
+            'test_benchmark_618' if args.test == 'benchmark' else 'test_remaining_534'}
+        return [r for r in manifest if r['component'] in components]
+    if args.resources:
+        return [r for r in manifest if r['component'] == 'benchmark_resources']
+    emd = args.emd.upper()
+    if not emd.startswith('EMD-'):
+        emd = 'EMD-' + emd
+    if emd not in entry_map:
+        raise ValueError(f'{emd} is not in the fixed v1.0 release.')
+    return [r for r in manifest if r['path'] == entry_map[emd]]
+
+def main():
+    parser = argparse.ArgumentParser(description=__doc__)
+    group = parser.add_mutually_exclusive_group(required=True)
+    group.add_argument('--part', type=int, choices=(1, 2, 3, 4), nargs='+')
+    group.add_argument('--test', choices=('benchmark', 'remaining', 'all'))
+    group.add_argument('--resources', action='store_true')
+    group.add_argument('--emd', help='Find and download the complete archive containing this EMD ID.')
+    parser.add_argument('--destination', type=Path, default=Path('CryoAtlas-downloads'))
+    parser.add_argument('--revision', default='main')
+    parser.add_argument('--dry-run', action='store_true', help='Show selected archives and sizes without downloading them.')
+    args = parser.parse_args()
+    # Pin all reads and downloads to one repository revision.
+    revision = HfApi().dataset_info(REPO, revision=args.revision).sha
+    def read_json(name):
+        return json.loads(Path(hf_hub_download(REPO, name, repo_type='dataset', revision=revision)).read_text())
+    manifest = read_json('catalog/archives.json')
+    entry_map = read_json('catalog/entry_archives.json') if args.emd else None
+    try:
+        selected = select_archives(manifest, args, entry_map)
+    except ValueError as exc:
+        parser.error(str(exc))
+    if not selected:
+        parser.error('No matching archives.')
+    print(f'Revision: {revision}')
+    for row in selected:
+        print(f"{row['bytes']/1e9:8.3f} GB  {row['path']}")
+    print(f"Total: {sum(r['bytes'] for r in selected)/1e9:.3f} GB")
+    if args.emd:
+        print('Selection is an entire archive containing this entry, not an individual-entry download.')
+    if args.dry_run:
+        return
+    for row in selected:
+        path = Path(hf_hub_download(REPO, row['path'], repo_type='dataset', revision=revision, local_dir=args.destination))
+        if path.stat().st_size != row['bytes']:
+            raise RuntimeError(f'Size mismatch: {row["path"]}')
+        sha = hashlib.sha256()
+        with path.open('rb') as stream:
+            for block in iter(lambda: stream.read(8 * 1024 * 1024), b''):
+                sha.update(block)
+        if sha.hexdigest() != row['sha256']:
+            raise RuntimeError(f'SHA-256 mismatch: {row["path"]}')
+        print(f'Verified: {row["path"]}')
+
+if __name__ == '__main__':
+    main()