SAVRN
Search Contact SAVRN

Organization

AnchorSR

AnchorSR

Models in Library0
Datasets in Library1
Models on Hugging Face9
Followers2

Datasets

Dataset · Visual question answering

TrainingData_Stage3

AnchorSR

This repository is a provenance-preserving collection of spatial measurement questions for answer-supervised training. It is built from VSI-590K, SpaceVista-Full, HiSpatial-Data, and CA-VQA. SenseNova-SI-8M is intentionally out of scope for this release. The first deliverable is the complete master collection. Smaller and larger training views will be derived only after all eligible metric examples have been retained and audited; they are not early sampling quotas. - annotations/ /measurement.parquet: canonical question/answer rows. - manifests/ /mediainventory.jsonl: unique required media and usage. - audits/ /: selection counts, validation, checksums, and source policy. - media/ /.tar…

Publicly accessible other