RelateAnything infers object relations in real time with only 53M parameters
RexDouglass · x · 2026-09-16
RelateAnything, a new model by Maëlic Neau, estimates relationships between objects in images in real time with just 53M parameters.
- Reasons over object regions from YOLO, RF-DETR, SAM and similar detectors/segmenters
- Trained on 474K images with over 4.3M relation annotations, covering 10,102 relation types
- Accepts specified relation words and generalizes to unseen relations without retraining
- Runs at roughly 20ms per frame
More from Research
- Cartesia explains why benchmarking TTS is extremely hard: no single number captures voice quality — saranormous · 2026-09-16
- New Paper on Inventor Positioning in Idea Space and What It Implies for Growth — danielrock · 2026-09-16
- Schmidhuber: LLMs lack creativity because they ignore his 2008 Fun theory — SchmidhuberAI · 2026-09-16
- DeepMind's guide to sharded matrix multiplication: the math behind training LLMs on 10k TPUs — zacharynado · 2026-09-16
- USC lecture notes derive closed-form AUC scaling laws and a 'spectral horizon' for finite-data ML — PTenigma · 2026-09-16
- Just having a tool available drops model answer rate from 98.2% to 63.5%, paper finds — dair_ai · 2026-09-16