Questions about Thermal Image

[Choo-chu]

Hello, I came across your repository through the paper “NUDT4MSTAR: A New Dataset and Benchmark Towards SAR Target Recognition in the Wild.”

I noticed that NUDT4MSTAR provides the same GitHub repository link as ATRNet-STAR, which led me to explore it further.

While there seem to be some differences between ATRNet-STAR and NUDT4MSTAR, I am particularly curious about whether the thermal images mentioned in NUDT4MSTAR are actually provided in the dataset.

If they are available, I would also like to know whether multi-view images (e.g., SAR and RGB) corresponding to the same targets are provided as well.

I truly appreciate your research and look forward to your reply. Thank you.

[waterdisappear]

Regarding your question about thermal images:

The NUDT4MSTAR/ATRNet-STAR dataset did collect corresponding thermal imaging data. However, the current release (v1) does not include these thermal images. The main reason is that we are still addressing the complex pixel-level alignment issues between multi-modal images (SAR, thermal, and RGB). Due to significant differences in imaging mechanisms, resolutions, and viewing angles across different sensors, achieving high-precision automatic alignment requires extensive manual verification and algorithmic optimization to ensure data quality meets the standards for academic research.

Therefore, I estimate that a version with fully aligned multi-modal images may take about 1–2 years to release.

If your research urgently requires multi-modal target data, I recommend checking out the SMART-Ship dataset from other research groups. This dataset has done excellent work on multi-modal alignment and may be a more mature option at present.

Thank you again for your attention and understanding! Please feel free to reach out if you have further questions.

[Choo-chu]

Thank you very much for your detailed explanation.

I really appreciate your clarification regarding the thermal images and the multi-modal alignment challenges. It is completely understandable that high-quality pixel-level alignment requires significant effort and time.

Thank you also for recommending the SMART-Ship dataset — I will definitely take a look at it.

I sincerely appreciate your kind response and wish you the best in your ongoing research.