README.md

metadata

license: apache-2.0
language:
  - en
task_categories:
  - image-text-to-text
size_categories:
  - 10K<n<100K
pretty_name: GSXray

GSXray

GSXray is a 44,019-example supervised fine-tuning corpus for dual-view X-ray reasoning. Each record contains a paired top-view and side-view image plus a two-turn human/assistant conversation. The public paths are relative and ordered as top then side.

Files

• annotations/sft_merged_44k.json: LLaMA-Factory-compatible array preserving the original conversation text.
• annotations/train.jsonl: stable line-oriented records with id, conversations, and images.
• data/train-NNNNN.tar: deterministic uncompressed image shards. Members are images/top/ID.jpg and images/side/ID.jpg.

Extract every tar shard into the dataset root before consuming the annotation paths. There are exactly 44,019 unique IDs and 88,038 image members.

License and provenance

Released under Apache-2.0 under the confirmed redistribution basis recorded for this release. GSXray is derived from the LDXray dual-view X-ray dataset and adds GSR annotations, structured reasoning data, and release packaging. See NOTICE and LICENSE.

Citation

@inproceedings{peng2026gsr,
  title={Can a Second-View Image Be a Language? Geometric and Semantic Cross-Modal Reasoning for X-ray Prohibited Item Detection},
  author={Peng, Chuang and Tao, Renshuai and Ren, Zhongwei and Liu, Xianglong and Wei, Yunchao},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  year={2026}
}

@inproceedings{tao2025ldxray,
  title={Dual-view X-ray Detection: Can AI Detect Prohibited Items from Dual-view X-ray Images Like Humans?},
  author={Tao, Renshuai and Wang, Haoyu and Guo, Yuzhe and Chen, Hairong and Zhang, Li and Liu, Xianglong and Wei, Yunchao and Zhao, Yao},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={10338--10347},
  year={2025}
}