MobileFetalCLIP is a highly efficient foundation model designed specifically for fetal ultrasound analysis on point-of-care, low-resource devices (like smartphones). It achieves this by distilling knowledge from a massive 427M parameter teacher model into a tiny 11.4M parameter student model using a novel technique called Selective Repulsive Knowledge Distillation.
Please refer to the official GitHub repository for installation instructions, dataset preparation, and inference scripts:
🔗
GitHub: numanai/MobileFetalCLIPIf you find this model or codebase useful for your research, please cite the paper:
1@article{saeed2026mobilefetalclip,
2 title = {MobileFetalCLIP: Selective Repulsive Knowledge Distillation
3 for Mobile Fetal Ultrasound Analysis},
4 author = {Saeed, Numan and Maani, Fadillah Adamsyah and Yaqub, Mohammad},
5 journal = {arXiv preprint arXiv:2603.05421},
6 year = {2026}
7}
