leabogensperger/joint-flow-matching-cells

Joint Flow Matching Enables Continuous Dose-Conditioned Cell Morphing

Pretrained checkpoints for the paper accepted at the ECCV 2026 Workshop on Bioimage Computing (BIC). Code, training/evaluation scripts, and full documentation: https://github.com/leabogensperger/joint-flow-matching-cells

Contents

vae/config.yaml, vae/best_model.pth              # shared microscopy VAE (all cell lines/drugs)
flow/<drug>/config.yaml, best_model.pt,
             latent_scale.pt                      # one joint flow matching model per drug

Download with scripts/download_checkpoints.py from the GitHub repo above; the files land directly in experiments/, ready to use with evaluate_vae.py / evaluate_latent_flow_joint.py.

Citation

@article{bogensperger2026joint,
  title={Joint Flow Matching Enables Continuous Dose-Conditioned Cell Morphing},
  author={Bogensperger, Lea and Merlo, Manuela and Baumgartner, Martin and Krauthammer, Michael and Ciraulo, Bernard},
  journal={arXiv preprint arXiv:2608.16424},
  year={2026}
}