In your paper you say: "Here, we first provide the description of notations used in
the following ablation study and tables. M(0) is the initial
model to start the bidirectional learning and is trained only
with source data." and "In our experiments, we choose to
use DeepLab V2 [3] with ResNet101 [11] and FCN-8s [18]
with VGG16 [32] as our segmentation model. They are initialized with the network pre-trained with ImageNet [15]."
So the Deeplabv2_init you made for download is trained on Source data or is it the model with ImageNet weights in the backbone?
It's quite important since I need to conclude experiments on my dataset, and I'd like to know if I need to train in on Source data or not.
In your paper you say: "Here, we first provide the description of notations used in the following ablation study and tables. M(0) is the initial model to start the bidirectional learning and is trained only with source data." and "In our experiments, we choose to use DeepLab V2 [3] with ResNet101 [11] and FCN-8s [18] with VGG16 [32] as our segmentation model. They are initialized with the network pre-trained with ImageNet [15]." So the Deeplabv2_init you made for download is trained on Source data or is it the model with ImageNet weights in the backbone? It's quite important since I need to conclude experiments on my dataset, and I'd like to know if I need to train in on Source data or not.