How you conducted testing on the Aachen dataset?

[Inverse-function]

Hello, when I integrated your code for testing on the Aachen Day-Night dataset for long-term visual localization, I encountered an error at the transformer stage indicating that two tensor sizes did not match. Does your code enable a feature similar to rectangular inference? Could you share how you conducted testing on the Aachen dataset?

[wyf2020]

Sorry for the late reply. Our code supports rectangular images, for example, the images from ScanNet are rectangular.