FDIF pretrains on formula-generated synthetic 3D data, hitting 81.89% without real medical images
HirokatuKataoka · x · 2026-10-31
FDIF, accepted to ACCV 2026 by researchers including AIST's Hirokatsu Kataoka, uses signed distance functions (SDFs) to synthesize 3D training data directly from mathematical formulas.
- Pre-training requires no real medical images or expert annotations, sidestepping privacy constraints and labeling costs.
- Unlike voxel-based FDSL (32 classes, no textures), FDIF supports 109 classes plus arbitrary implicit functions and synthesizes both geometric and intensity textures via distance fields.
- Across AMOS, ACDC, and KiTS with SwinUNETR, nnUNet ResEnc-L, and nnUNet Primus-M, it consistently beats prior formula-driven methods, reaching 81.89% with SwinUNETR — comparable to self-supervised pre-training on large real datasets.
More from Research
- New paper proposes the Piggyback Hypothesis: chat-template tokens carry emergent misalignment out of domain — davidbau · 2026-10-31
- FPV-SLAM claims 1st place on ETH Zurich's city-scale LaMAria mono-inertial benchmark — ravithejads · 2026-10-31
- FDSML (WACV 2027): fractal-parameter metric learning lifts image retrieval up to +22pt — HirokatuKataoka · 2026-10-31
- Early OpenAI interpretability researcher: most neurons in trillion-param models are now legible — CatAstro_Piyush · 2026-10-31
- Image-Space Rule Discovery paper accepted to ACCV 2026, updated arXiv version released — HirokatuKataoka · 2026-10-31
- Measured: lossy engine flags drop a Q8_0 model to roughly Q2_0 quality — jesdga95 · 2026-10-31