train my own dataset

[HHEjie123]

can you tell me how to train on my own dataset,What is the dataset format like?

[chundiliu]

Hi HHEjie123,

We followed GeM model to generate the descriptors and evaluate the performance.

Our code requires a [Nxd] descriptor matrix as the input. The code will build the kNN graph automatically.

By doing a forwardpass through the pretrained GeM model, you shoul be able to get the global descriptors for your own dataset. Loading the descriptors and your own ground-truth files into our pipeline should be straight foward.

Please let us know if you have any other issue running the code with your own datasets.