haohxin/cs396-final

CS396 Final Dataset (Class-Medium Real Vision)

This dataset snapshot contains:

• pilot_3k_class_medium_real_vision.jsonl (21000 rows)
• vision_corrupt/class_medium/pilot_3k/ (9000 materialized occlusion images)
• manifests/configs/scripts required for deterministic reconstruction

Reproducibility

Primary manifest: pilot_3k_class_medium_real_vision.manifest.json

• output JSONL SHA256: 42e677db5c99c657ac44946749c0c800c06df4ec7fe9d8fc919ac94c7e92b990
• image-dir fingerprint SHA256: fc835c8f7e7aa24f6b22007f8596f3f24a5e28f58841b5aecbf2282d1b873123

Notes

• Source data comes from VQAv2 + COCO.
• This package includes generated artifacts for collaboration convenience.