LocateAnything-3B randomized-query delivery detector R2
Personal, non-commercial LocateAnything continual-SFT experiment.
Starting model: davidr99/locateanything-3b-delivery-detector-lora
Classes: Amazon, UPS, USPS-Truck, Other-Vehicles, FedEx
R2 continues from the successful full-list R1 detector. Each image is trained with a deterministic randomized full-list, positive-subset, single-positive, or absent-class-negative query. The answer contains boxes only for requested classes, and class order is randomized.
Input images are capped at 640px on the longest side. Coordinates are normalized integers in the range 0 to 1000.
Prompt seed: 20260803
Query mixture: {"full": 0.2, "negative": 0.15, "positive_subset": 0.45, "single_positive": 0.2}
License
This model is a derivative of NVIDIA LocateAnything-3B and remains restricted to non-commercial research/evaluation under NVIDIA's model license.
