Dataset card: describe FluXray inline

README.md

@@ -15,27 +15,17 @@ This repository holds every real X-ray source whose license permits redistributi
 repackaged in FleXray's common label protocol in the `fxr-dataset` package format.
 Two companion releases complete the data described in the paper:
 
-- **FluXray** (synthetic, CC BY-NC 4.0): `VictorButoi/fluxray`
+- **FluXray** (synthetic, CC BY-NC 4.0): `VictorButoi/fluxray` — 138,063 generatively
+  edited digitally reconstructed radiographs rendered from the 1,597 MOOSE CTs at 90 poses
+  each, quality-filtered, with exact overlapping masks for all 63 protocol structures.
+  Shipped as the training database itself (`FluXray/thunder_dbs/7.1/data.mdb`, an LMDB
+  of float16 256 x 256 images and 63-channel binary masks that `flexray` reads directly)
+  together with `samples.csv` (pose, split, MOOSE subject and per-sample license for
+  every image), `protocol.yml` (label order and mask thresholds), and the
+  `filter_*.csv` quality-control scores and thresholds.
 - **MURA forearm/humerus annotations** (masks only, CC BY 4.0): bundled with the
   FleXray GitHub repository and mirrored here under `mura_forearm_humerus_annotations/`
 
-### What is in `VictorButoi/fluxray`
-
-138,063 generatively edited digitally reconstructed radiographs rendered from the 1,597
-MOOSE / ENHANCE-PET CTs (90 camera poses each, ~4% removed by quality control), with
-exact overlapping masks for all 63 protocol structures. Unlike the folders here it is
-not a `fxr-dataset` package but the training database itself, laid out as
-`FluXray/thunder_dbs/7.1/` so that `flexray` reads it directly
-(`export GENERATED_DATAPATH=<download>`; config `FluXray: {version: "7.1"}`):
-
-| File | Contents |
-|---|---|
-| `data.mdb` (18.8 GB) | LMDB with one entry per sample: `img` (float16 256 x 256 in [0, 1]) and `seg` (uint8 63 x 256 x 256 binary channel masks, channels may overlap), plus `_samples`, `_splits`, `_metadata` and `_attrs` index entries. |
-| `samples.csv` | One row per sample: MOOSE subject, split, pose id, camera position and rotation, number of visible structures, source site and per-sample license. |
-| `protocol.yml` | The 63 label names in channel order and the mask-binarization thresholds. |
-| `filter_kept.csv`, `filter_dropped.csv`, `filter_thresholds.csv` | Quality-control scores (soft Dice of a reference segmenter on the edited vs. unedited render) and the per-pose thresholds that kept 138,064 of 143,730 renders. |
-| `README.md`, `LICENSE` | Card and license: CC BY-NC 4.0 overall; 50,282 samples from CC BY 4.0 CTs (Leipzig, Careggi) and 87,781 from CC BY-NC 4.0 CTs (AutoPET), recorded per sample. |
-
 ## Redistributed datasets (CC BY 4.0)
 
 | Dataset | Role in paper | Samples (train/val/test) | Labels | Source |