Rewrite model card: clearer docs, dataset links, no checkpoint iters

README.md

@@ -8,71 +8,78 @@ tags:
   - registration
   - dynamic-mri
   - cmrxrecon
+pipeline_tag: image-to-image
 ---
 
 # DIRECT — Adaptive Sampling, Reconstruction & Registration
 
-Pretrained checkpoints and matching YAML configs for
-[Deep End-to-End Adaptive k-Space Sampling, Reconstruction, and Registration for Dynamic MRI](https://arxiv.org/abs/2411.18249)
-(Yiasemis et al.).
+Pretrained models that jointly learn adaptive \(k\)-space sampling, reconstruction, and motion registration for dynamic MRI.
+
+- Paper: [arXiv:2411.18249](https://arxiv.org/abs/2411.18249)
+- Companion (sampling + recon only, MIDL 2026): [PMLR](https://proceedings.mlr.press/v315/yiasemis26a.html) · Hub [`NKI-AI/direct-e2e-ads-recon`](https://huggingface.co/NKI-AI/direct-e2e-ads-recon)
+- Framework: [DIRECT](https://github.com/NKI-AI/direct)
 
 ## Dataset
 
 | | |
 |---|---|
-| **Dataset** | [CMRxRecon](https://cmrxrecon.github.io/) cine (cardiac MRI) |
-| **Domain** | Multi-coil dynamic / cine MRI |
-| **Task** | Joint adaptive sampling + reconstruction + registration |
-
-## Papers
+| **Data** | [CMRxRecon](https://cmrxrecon.github.io/) multi-coil cardiac cine |
+| **Challenge / site** | [cmrxrecon.github.io](https://cmrxrecon.github.io/) |
+| **Task** | Adaptive sampling + reconstruction + registration |
 
-- **This work:** [arXiv:2411.18249](https://arxiv.org/abs/2411.18249)
-- Companion (sampling + recon only, MIDL 2026): [PMLR](https://proceedings.mlr.press/v315/yiasemis26a.html) · models in [`NKI-AI/direct-e2e-ads-recon`](https://huggingface.co/NKI-AI/direct-e2e-ads-recon)
+## Models
 
-## Files
-
-```text
-<name>.yaml
-<name>.pt
-```
+Each experiment is a `.yaml` + `.pt` pair:
 
 | Name | Notes |
-|---|---|
-| `vsharp_ads_1d_phase_reg` | vSHARP + ADS phase-specific + registration (E2E) |
+|------|-------|
+| `vsharp_ads_1d_phase_reg` | vSHARP + ADS phase-specific + registration (end-to-end) |
 | `varnet_ads_1d_phase_reg` | VarNet + ADS phase-specific + registration |
 | `vsharp_ads_1d_reg` | vSHARP + ADS unified + registration |
 | `varnet_ads_1d_reg` | VarNet + ADS unified + registration |
-| `vsharp_ads_1d_phase_init_reg` | phase-specific + sampling init |
-| `vsharp_ads_1d_init_reg` | unified + sampling init |
-| `*_disjoint` | stage-wise / disjoint training (`train_end_to_end: false`) |
-| `vsharp_fixed_1d_*` | fixed (non-adaptive) sampling baselines |
-| `vsharp_loupe_1d_*` | LOUPE / optimized sampling baselines |
+| `vsharp_ads_1d_phase_init_reg` | Phase-specific + sampling init |
+| `vsharp_ads_1d_init_reg` | Unified + sampling init |
+| `*_disjoint` | Stage-wise training (`train_end_to_end: false`) |
+| `vsharp_fixed_1d_*` | Fixed (non-adaptive) sampling baselines |
+| `vsharp_loupe_1d_*` | LOUPE / optimized-sampling baselines |
 
-## Training protocol & inference masking
+Full training configs: [`projects/e2e_ads_recon_reg`](https://github.com/NKI-AI/direct/tree/main/projects/e2e_ads_recon_reg).
 
-Same domain as the companion E2E-ADS-Recon models: **CMRxRecon** cine with
-mixed discrete accelerations (typically `[4.0327, 6, 8.2]` or init variants)
-and ACS `center_fractions` of matching length (usually `0.04`).
+## Training protocol
 
-Released inference YAMLs pin **one** R and **one** ACS (default `val-4x`).
-To change rate at inference, set a single pair under
-`inference.dataset.transforms.masking` (lists of length 1 only).
+Same data domain as the companion E2E-ADS-Recon models: CMRxRecon cine with mixed discrete accelerations (typically \(R \in \{4.0327, 6, 8.2\}\), or init variants) and ACS `center_fractions` of matching length (usually `0.04`).
+
+Released inference YAMLs pin **one** \(R\) and **one** ACS (default `val-4x` for that config). To change rate at inference, edit `inference.dataset.transforms.masking` and keep both lists length 1:
+
+```yaml
+masking:
+  name: FastMRIEquispaced
+  accelerations: [8.2]
+  center_fractions: [0.04]
+```
 
 ## Usage
 
 ```bash
+pip install direct-recon   # or: https://github.com/NKI-AI/direct
 hf download NKI-AI/direct-e2e-ads-recon-reg --local-dir ./e2e_ads_recon_reg
 
 direct predict ./predictions \
   --cfg ./e2e_ads_recon_reg/vsharp_ads_1d_phase_reg.yaml \
   --checkpoint ./e2e_ads_recon_reg/vsharp_ads_1d_phase_reg.pt \
-  --data-root /path/to/cmrxrecon/data \
+  --data-root /path/to/cmrxrecon \
   --num-gpus 1
 ```
 
-The first argument is the **output directory** (predictions), not an experiment folder.
+The first argument is the **prediction output directory**.
 
+## Citation
 
-## Framework
-
-[DIRECT](https://github.com/NKI-AI/direct) · [NKI-AI](https://huggingface.co/NKI-AI)
+```bibtex
+@article{yiasemis2024e2eadsreg,
+  title={Deep End-to-End Adaptive k-Space Sampling, Reconstruction, and Registration for Dynamic {MRI}},
+  author={Yiasemis, George and Moriakov, Nikita and Sonke, Jan-Jakob and Teuwen, Jonas},
+  journal={arXiv preprint arXiv:2411.18249},
+  year={2024}
+}
+```