Domen Tabernik

Hi, I have updated the website, you can find it under "download here" link.

Hi, the [prepared dataset file for TensorFlow](http://box.vicos.si/skokec/gostop/KolektorSDD-dilate=5-tensorflow.zip) already has dilation=5 included. To get other dilations, you can download the original png images from [here](https://www.vicos.si/Downloads/KolektorSDD) and then use [this script](https://github.com/skokec/segdec-net-jim2019/blob/master/input_data/input_data_build_image_data_with_mask.py) to...

Hi, thanks @onlyNata for explained it quite well. The purpose is NOT segmentation but image level classification. Segmentation is only used as auxiliary step to get good features for classification,...

Hi, we have not experience issues with the exploding gradients. Since each feature layer is normalized to 0-mean and unit scale there should not be any problems with that. You...

Hi, we only used GPU (never tested CPU actually), so it should work well with GPUs. It looks like you may not have correct drivers installed and TensorFlow does not...

Hi, For use in Keras you need to use the class version `DAUConv2dTF(filters, dau_units, max_kernel_size)`, which only works as `padding = same` and cannot be changed. There is actually no...

That is correct. Best, Domen

You need to store the data as TFRecords using the tf.train.Example. The required features that need to be added into the tf.train.Example can be seen in [input_data/image_processing.py](https://github.com/skokec/segdec-net-jim2019/blob/2ebe7f5e4080f3c6f1d5117ff534583c20b97308/input_data/image_processing.py#L337). I have also...

Hi, sorry for late reply as I wasn't able to find time before. Do you actually get an error that it cannot go further? It appears that there seems to...

Hi, there are two output nodes, one for segmentation that uses 'conv5' scope name and another for classification that uses 'decision/FC9' scope name. You can see segmentation node [here](https://github.com/skokec/segdec-net-jim2019/blob/694f7787df70d439fd722600cccd714a7e5c5732/segdec_model.py#L118) and...