weizmann-sampl: sync data card and figures with GitHub

Syncs the weizmann-sampl card and figures on the Hub with the reviewed state on GitHub (open-h/OpenH-RF).

One correction to the field table: `raw_data`'s axial dimension is 2176, not 4096 — and 2048 in one of the 30 acquisitions. The table now states both.

Adds the three figures. `assets/30_1.gif` is the cine loop the card now leads with, reconstructed from `data/30_1.hdf5` with the `pipeline.yaml` in this folder. `assets/three_patients_grid.png` is the validation figure — three frames from each of three subjects — which the card embedded from the folder root, where it was never uploaded, so it was broken here; its caption now describes the figure instead of a `reconstruct.py` invocation that no longer matches the script. `assets/30_1.png` is a single frame from the loop and is not embedded in the card.

Files changed: 4.

```
+ weizmann-sampl/assets/30_1.gif
+ weizmann-sampl/assets/30_1.png
+ weizmann-sampl/assets/three_patients_grid.png
~ weizmann-sampl/README.md
```

weizmann-sampl/README.md

@@ -14,6 +14,12 @@ tags:
 
 # OpenH-RF Thyroid Clinical Scans
 
+![Reconstructed cineloop from 30_1.hdf5](assets/30_1.gif)
+
+Cine loop of [`30_1.hdf5`](https://huggingface.co/datasets/nvidia/OpenH-RF/blob/main/weizmann-sampl/data/30_1.hdf5), reconstructed from the raw
+channel data with the `pipeline.yaml` in this folder.
+
+
 ## Dataset Description
 
 The data set consists of clinical ultrasound channel data acquired
@@ -69,7 +75,7 @@ the frames prior to workspace parameter freezing were removed from the raw chann
 
 | Field | Shape                        | Dtype | Units | Description |
 |---|------------------------------|---|---|---|
-| `tracks/track_0/data/raw_data` | (nframes, 128, 4096, 128, 1) | int16 | ADC counts | Raw per-element RF channel data: (frames, transmits, axial samples, elements, 1) |
+| `tracks/track_0/data/raw_data` | (nframes, 128, 2176, 128, 1) | int16 | ADC counts | Raw per-element RF channel data: (frames, transmits, axial samples, elements, 1); the axial dimension is 2176 in 29 of the 30 acquisitions and 2048 in the remaining one |
 | `tracks/track_0/scan/t0_delays` | (128, 128)                   | float32 | s | Per-transmit, per-element transmit delay |
 | `tracks/track_0/scan/tx_apodizations` | (128, 128)                   | float32 | unitless (0-1) | Per-transmit, per-element transmit apodization weight |
 | `tracks/track_0/scan/polar_angles` | (128,)                       | float32 | rad | Per-transmit polar steering angle |
@@ -110,11 +116,10 @@ apply window → demodulate → DAS beamform (with native, per-element/per-pixel
 lens correction) → envelope detect → normalize → log compress. See that file
 for the grid size, dynamic range and lens-correction parameters.
 
-Example reconstruction, run via `python reconstruct.py --data-dir subjects
---n-scans 3 --n-frames 3 --seed 0 --output three_patients_grid.png`
-(3 subjects, 3 randomly sampled frames each):
+Three frames from each of three subjects, reconstructed with the `pipeline.yaml`
+in this folder:
 
-![3 frames from 3 different subjects](three_patients_grid.png)
+![3 frames from 3 different subjects](assets/three_patients_grid.png)
 
 All 9 sampled frames, across 3 different subjects, show consistent diffuse
 in-vivo tissue speckle with no reconstruction artifacts, confirming the