drift-hcp

DRIFT — HCP pretrained checkpoints

Pretrained checkpoints for DRIFT: Difficulty-aware Rectified Flows for Through-plane MRI Super-Resolution (ECCV 2026), trained on the HCP dataset (0.7 mm isotropic T1w/T2w).

• drift_hcp_stage1.ckpt — Stage 1: Anatomical Projection Network (APN)
• drift_hcp_stage2.ckpt — Stage 2: Difficulty-aware Rectified Flow

Checkpoints are weights-only (optimizer state stripped) for lightweight inference.

Usage

from models.drift_2d import Stage1APNLightning, Stage2RFLightning
stage1 = Stage1APNLightning.load_from_checkpoint("drift_hcp_stage1.ckpt", map_location="cpu")
stage2 = Stage2RFLightning.load_from_checkpoint("drift_hcp_stage2.ckpt", map_location="cpu")

See the code repository: https://github.com/yoonseokchoi-ai/DRIFT
Project page: https://yoonseokchoi-ai.github.io/drift-eccv2026/

Citation

@inproceedings{drift2026,
  title={DRIFT: Difficulty-aware Rectified Flows for Through-plane MRI Super-Resolution},
  author={Choi, Yoonseok and Ha, Eun-Gyu and Kim, Daniel and Al-masni, Mohammed A. and Yang, Ming-Hsuan and Kim, Dong-Hyun},
  booktitle={European Conference on Computer Vision (ECCV)},
  year={2026}
}