How to test the joint train segmentation network?

[linyangfei]

Hello! I have read your paper and want to train the joint segmentation model. But I have some problems about the scripts. In your "train_seg.prototxt", you joint the denoising network and segmentation network together. When running "main_train_seg.sh":
python train_net_joint.py --solver=config/solver_seg.prototxt --weights=model/denoise-s30.caffemodel --weights_2=model/seg_init.caffemodel --GPU=0

two sets of weights(denoising network and segmentation network) into the model. In "main_test_seg.sh":
../build/tools/caffe test --model=config/test_seg.prototxt --weights=model/seg_joint.caffemodel --gpu=0 --iterations=1449

one set of weights(seg_joint.caffemodel) is parsed into the model, but the test_seg.prototxt only contains the segmentation network, not the joint network. So when I run the script, the joint weights of seg_joint.caffemodel cannot parsed into the single segmentation model. Could you tell me how to evaluate the joint trained model?
Looking forward to your reply. Thank you!