Joel Akeret

Why do you think this is happening? In line [98](https://github.com/jakeret/tf_unet/blob/master/tf_unet/unet.py#L98) the `in_node` gets overwritten in every iteration

No worries. I had a quick look and I think things get a bit tricky as soon as one uses a filter size > 3. It's certainly doable but I...

On thing that might be worth to look into is image normalization. If your dataset is unbalance it could help to over/undersample the data. Another thing you could try is...

If I understand correctly you have 3 classes: settlements / slums / background. Hence, I would try to train a model that is able to distinguish settlements from anything else...

You could write your own DataProvider by subclassing the `ImageDataProvider` and overwriting the [`_load_file` function](https://github.com/jakeret/tf_unet/blob/master/tf_unet/image_util.py#L178)

The error message seems to be missing. There is only the traceback I'm also not quite sure what you changed. This looks like an entire copy of the tf_unet image_util.py..

Possible that there is a bug in the [upconv](https://github.com/jakeret/tf_unet/blob/master/tf_unet/layers.py#L44) or in the [concatenation](https://github.com/jakeret/tf_unet/blob/master/tf_unet/layers.py#L55). Would be great if you could try to adapt the current implementation to see if is solves...

I'm not sure where the bug is. Probably you have to step through the code and look at the tensor shapes

The number of objects in the image doesn't matter. You can have the same class multiple times Looks like the [ImageDataProvider](http://tf-unet.readthedocs.io/en/latest/tf_unet.html#tf_unet.image_util.ImageDataProvider) s what you need. Besides the images and masks...

Have you experimented with the network architecture (# of layers, # of filters)?