I am trying to implement Deeplab model for semantic segmentation in Caffe2. I tried to use SigmoidCrossEntropyWithLogits as the loss function but I got this warning:
WARNING:root:You are creating an op that the ModelHelper does not recognize: SigmoidCrossEntropyWithLogits.
And it seems that the training process does not run correctly.
Does anybody try to implement a semantic segmentation (pre-trained on ImageNet) code in Caffe2?
Hi,
I am trying to implement Deeplab model for semantic segmentation in Caffe2. I tried to use SigmoidCrossEntropyWithLogits as the loss function but I got this warning:
WARNING:root:You are creating an op that the ModelHelper does not recognize: SigmoidCrossEntropyWithLogits.
And it seems that the training process does not run correctly.
Does anybody try to implement a semantic segmentation (pre-trained on ImageNet) code in Caffe2?
Thanks.