Questions about the pretrained model

[missyoudaisy]

Hello, I read your code and paper in recent days. However, I have one question. what does the pretrained model you share corresponds to ? The pixel-level, the feature-level or the response level? Cuz the input channels of models are different. For example, the input channel of pixel-level is 4, while the input channel of feature-level and response-level is 3.