ONNX-exported and quantised vision-language models for Scene Atlas. These are derivative works — see individual model cards for attribution.