In this site there are labeled satellite image datasets for semantic segmentation but those images are not multi-temporal. In this question someone asks for satellite images but again, he doesn't ask for multi-temporal data. Where can I find labeled (as in, categorized, see here for an example), open datasets for multi-temporal remote sensing image analysis, more specifically semantic segmentation and deep learning?

Applications could be anything like land use, land cover segmentation; change detection, deforestation tracking or others. Multi-spectral or hyper-spectral images would be preferred. SAR, LiDAR or images from UAVs would also work.

  • What kind of labels?
    – csk
    Jun 5, 2018 at 20:54
  • It could be crop labels. Example: an image with each pixel being red if the crop type in that pixel location is soybean, yellow if the crop type is corn, orange if it is grain, etc. This would be the desired values for the supervised classification task. Jun 6, 2018 at 0:27
  • Here and here are some label examples for semantic segmentation Jun 8, 2018 at 9:29
  • The images you linked to are both the same image, and neither of them is labeled. "Label" means text. It sounds like you're referring to categorized images.
    – csk
    Jun 12, 2018 at 1:48


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.