DeadAhead: fine-tuned VideoMAE models and survival heads

The models behind the paper DeadAhead: Forecasting when and how a cell will die from quantitative phase video, one per held-out group and seed: 27 fine-tuned VideoMAE-L classifiers and 27 DeadAhead survival heads. A model whose name carries a group was trained without that group. Code, model card, file sizes and sha256: https://github.com/Mahdi-Koochali/DeadAhead, docs/MODEL_CARD.md and docs/WEIGHTS.md. Licences. videomae_large_finetuned__*.safetensors: CC BY-NC 4.0, the licence of the checkpoint they start from (MCG-NJU/videomae-large; Tong et al., NeurIPS 2022, arXiv:2203.12602). deadahead_head__*.txt: MIT; their inputs include the features of those models.