Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note2 min read

JL1-CC&QA: Extending the JL1-CD Benchmark with Change Captioning and Question Answering

Source

arXiv

arxiv.org/abs/2606.31745v1

Read the source

Source

Plain-Language Brief

Researchers introduce JL1-CC&QA, a new benchmark for remote sensing that extends the JL1-CD dataset with two annotation layers: change captioning (CC) and change question answering (QA). The benchmark includes 5,000 bi-temporal image pairs from the Jilin-1 satellite, providing 17,021 quality-verified captions describing diverse land-cover transformations and 20,060 question-answer pairs across eight question types. The annotations are generated through a three-stage pipeline involving multiple large language models, vision-grounded language models, and human experts.

What the Document Says

  • The JL1-CC&QA benchmark is introduced for remote sensing.
  • It extends the JL1-CD dataset with change captioning (CC) and question answering (QA) annotation layers.
  • There are 5,000 bi-temporal image pairs from the Jilin-1 satellite in the benchmark.
  • JL1-CC provides 17,021 quality-verified captions for diverse land-cover transformations.
  • JL1-QA offers 20,060 question-answer pairs across eight question types.
  • The annotations are produced through a three-stage pipeline involving multiple large language models, vision-grounded language models, and human experts.

Key Concepts Mentioned

  • remote sensing
  • change detection
  • JL1-CD dataset
  • change captioning
  • change question answering
  • annotation layers
  • Jilin-1 satellite
  • bi-temporal image pairs
  • large language model
  • vision-grounded language model

Claims Made by the Document

  • The JL1-CC&QA benchmark will advance multi-task change understanding in remote sensing.

Evidence Used

  • All annotations are produced via a three-stage pipeline consisting of multi-modal large language model (LLM) generation, vision-grounded LLM judging, and human expert verification. The dataset is available at https://github.com/circleLZY/JL1-CD.

Methods / Approach

  • Three-stage pipeline consisting of multi-modal large language model generation, vision-grounded language model judging, and human expert verification

Limitations Stated by the Authors

  • UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

  • The research paper does not explicitly mention limitations. However, as with any dataset or methodology, there may be potential limitations related to the quality of annotations, generalizability across different datasets, and potential biases.

Open Questions

  • Potential questions could include: What are the specific eight question types in JL1-QA? How does the three-stage pipeline work in detail? What is the ground truth quality of the captions provided by JL1-CC?

Notes for Future Q&A

May include interpretation. Labeled where it does.

  • The research paper mentions eight question types, but it does not specify what they are.
  • The details of the three-stage pipeline for annotation generation are not explained in depth.

Raw Abstract / Excerpt

Remote sensing change detection (CD) traditionally focuses on pixel-level binary segmentation, which identifies where changes occur but neither what nor why. To bridge this semantic gap, we introduce JL1-CC&QA, a multi-task benchmark that extends the JL1-CD dataset with two complementary annotation layers: change captioning (CC) and change question answering (QA). Built upon 5,000 bi-temporal image pairs acquired by the Jilin-1 satellite at 0.5-0.75m ground sample distance, the benchmark comprises: (i) JL1-CC, providing 17,021 quality-verified captions that describe diverse land-cover transformations; and (ii) JL1-QA, offering 20,060 question-answer pairs across eight question types, enabling fine-grained, interactive interrogation of surface changes. All annotations are produced via a three-stage pipeline consisting of multi-modal large language model (LLM) generation, vision-grounded LLM judging, and human expert verification. We hope that JL1-CC&QA, as a benchmark unifying binary change masks, change captions, and change-oriented QA over the same image set, will serve as a valuable resource for the community to advance multi-task change understanding in remote sensing. The dataset is available at https://github.com/circleLZY/JL1-CD.