Performing inference with CPU

[alexanderDuenas]

thanks for your code!

I have two questions:

1. Its possible to perform inference just with CPU?
2. Why in the inference process the support images are also used. In the code the outputs are: outputs = model(img, supp_class_ids=support_class_ids_final, category_codes=all_category_codes_final) its necessary the arguments supp_class_ids and category_codes.?

[ZhangGongjie]

1. No, the codebase does not support that. But I think it should be pretty easy to implement that by yourself.

2. Because for inference, we still need prototypes for all classes. During inference, all_category_codes_final are computed by averaging all support instances for each of the given class.