icoz69
icoz69
@JinglunYang hi, when i run in your branch, i still got the warning "dim should not be larger than 23". how can i solve it?
hello. are you using the raw code and dataset? its not trained with Imagenet.
hi, please see the data list and dataset loader
hi. i only upload the one-shot model here. as to the k-shot loader and network, i will clean the codes and upload soon. before that you easily implement them based...
hi thank your for your interests in our work. 1. this might be a bug in QPTH solver and I did not know the reason for now. 2. basically, you...
hi, can you specify which datasets
hi, please try DeepEMD-FCN by setting appropriate pyramid size, such as the combination of,e.g., [1,2,3]. It will have consistently better performance.
I forget the optimal pyramid structure for FCN. You can try multiple sizes, such as [1,2,5] [1,3,5] [1,2,3]...
by the way, the original size, i.e. 5, is already included in the pyramid, see [here](https://github.com/icoz69/DeepEMD/blob/bda94126b0a79fa8f8106fa98cfd2200c378e809/Models/models/Network.py#L195) You may delete this line and manually add 5 to set a more flexible...
Hi, try training opencv solver with a slight drop of performance.