FDSML (WACV 2027): fractal-parameter metric learning lifts image retrieval up to +22pt
HirokatuKataoka · x · 2026-10-31
Researchers behind FractalDB propose FDSML, accepted to WACV 2027: formula-driven supervised metric learning that uses fractal generation parameters as supervision instead of discrete labels. With FineFractalDB and a Parameter-Aware Triplet loss over IFS parameter space, ResNet-50 Recall@1 improves from 65.6% to 74.0% on Stanford Online Products and from 56.2% to 78.4% on Stanford Cars over FDSL pre-training — all without real images, also improving robustness under corruptions.
More from Research
- Chain of Thought conflates interpretability and faithfulness — two very different things — rao2z · 2026-10-31
- Qwenputer: a full RISC-V CPU compiled into Qwen 3.5-4B weights, even running another LLM — CatAstro_Piyush · 2026-10-31
- Claude-designed proteins bind four scorpion toxins: 318,484 candidates yield 17 hits — jiqizhixin · 2026-10-31
- Stanford CS 312 grades on predicting experiment outcomes, assignments worth 0% — yen_chen_lin · 2026-10-31
- François Fleuret invents a compact notation for attention tensor ops — francoisfleuret · 2026-10-31
- Weekend watch: 2 hours on post-training agents with open models — SergioPaniego · 2026-10-31