AI2-THOR kitchen perception models: FloorPlan1
Status: public release, part of MISE (project page; an English paper is in preparation). Part of an MSc thesis (University of Tehran): an LLM + RAG embodied agent that plans L1 atomic / L2 sequential / L3 abstract kitchen commands in AI2-THOR, with an honest perception stack (no simulator ground truth at test time).
FloorPlan1 is the development kitchen; the numbers below are its registered perception evaluation (sealed holdout opened once).
Models
| File | Role |
|---|---|
FloorPlan1.yolo26s.pt |
primary detector, YOLO26s, 49 type-only classes, every frame |
FloorPlan1.amb26m.pt |
look-alike specialist, YOLO26m (Fork, Spoon, ButterKnife, Knife, Spatula, SaltShaker, PepperShaker, Plate, Pot, Pan); replaces the primary's predictions for those classes only while the agent holds an object up (gated) |
FloorPlan1.yolo26s.control.pt |
control: YOLO26s fine-tuned on primary + specialist data (the comparison the specialist is tested against) |
FloorPlan1.state_*.pt |
7 state classifiers (broken, cooked, dirty, filled, on, open, sliced), YOLO11s-cls |
SHA256SUMS lists the exact deployed bytes (they match the frozen Kitchen 1 hashes registered in amendment A38).
Results
| Measure | Value |
|---|---|
| primary detector (YOLO26s), TEST / CLOSE-TEST / LIVE mAP50 | 0.907 / 0.913 / 0.927 |
| primary detector, TEST / CLOSE-TEST / LIVE mAP50-95 | 0.803 / 0.821 / 0.811 |
| sealed holdout: held-target naming success, primary / gated (with specialist) / control | 0.483 / 0.924 / 0.791 |
Every number belongs to this kitchen's own adapted detectors and is never pooled across kitchens.
Licence and credits
AGPL-3.0 (fine-tuned from Ultralytics YOLO models). Training images rendered with AI2-THOR (Apache-2.0).
- Downloads last month
- 213