| SIAM: Segment It All Model, v0.3 | |
| Original authors: R. Valabregue, I. Khemir, E. Bardinet, F. Rousseau, G. Auzias, R. Dorent | |
| Original repository: https://github.com/romainVala/SIAM (Apache-2.0) | |
| Original weights: https://zenodo.org/records/17733431 (pred_DS108_LcsfP_Ano, CC-BY-4.0) | |
| Paper: https://arxiv.org/abs/2605.02737 | |
| The weights are licensed under CC-BY-4.0 (https://creativecommons.org/licenses/by/4.0/). | |
| Credit the original authors and cite the paper above. | |
| Changes: the PyTorch nnU-Net checkpoints were converted to float32 MLX safetensors | |
| for nnunet-mlx. Optimizer state and duplicate decoder aliases of encoder weights were | |
| not carried over. No retraining, pruning or quantization. | |
| MLX conversion by Joseph Sandoval and Alfredo Sandoval Jr. | |