Marvin Teichmann

The code is up to date. This code has been used for the experiments.

Did you initialize all submodules? Try `git clone --recursive [email protected]:MarvinTeichmann/MultiNet.git` I am not having this issue, I just double checked. Cloned MultiNet into a fresh virtuelenv and it works fine.

ConvCRFs can be trained using [PyTorch](https://pytorch.org/tutorials/beginner/pytorch_with_examples.html). Training is straight forward and can be done like any other neural network. Iterate over the training data, apply softmax cross-entropy loss and use...

Hi, it should be possible. I dont know much about the text sequence labelling problems, but my model follows closely the ideas of fully-connected CRFs. If you can apply fully-connected...

Hi, thanks for your interest in my work. Setting trainable to true and then backprop the gradients through the CRF is all which should be needed to train the CRF....

Hi Alex, the performance does not improve past 13. Also, if you go bigger then 21 (I think it was 21) I ran into GPU memory issues (@ 11GB). The...

Seams to be a python 3.6 / 2.7 compatibility error. If there is a way to make it work under both versions, feel free to create a pull request.

This is similar to #11. I use logsoftmax in the first iteration, because the original FullCRF implemented uses log on the data (followed by softmax). Logsoftmax does the same but...

hi @glhfgg1024, thanks for getting in touch with me. My interpretation is that the model adapts very quickly to the added CRF. The train mIoU is definitely not constant after...

Firstly I am using titan x (pascal) to measure runtime. K40 is rather old, so you might not get the same results. In addition, demo.py is not meant to measure...