Add seperate docu folder (#10)

* Update docu

* Update FAQ.md

README.md

@@ -80,7 +80,7 @@ Use the flag `--nosave` if you do not want to save all output in an rundir. This
 
 ### Modifying Model & Train on your own data
 
-The model is controlled by the file `hypes/KittiSeg.json`. Modifying this file should be enough to train the model on your own data and adjust the architecture according to your needs. A description of the expected input format can be found [here](inputs/inputs.md). I would advise to creat a new hype file `hypes/my_hype.json` for your input data and start trainining by running: `python train.py --hypes hypes/my_hype.json`
+The model is controlled by the file `hypes/KittiSeg.json`. Modifying this file should be enough to train the model on your own data and adjust the architecture according to your needs. A description of the expected input format can be found [here](docu/inputs.md). I would advise to creat a new hype file `hypes/my_hype.json` for your input data and start trainining by running: `python train.py --hypes hypes/my_hype.json`
 
 
 
@@ -138,6 +138,10 @@ In addition the following TensorVision environment Variables will be useful:
 
 On a cluster it is useful to set `$TV_USE_GPUS=force`. This will make the flag `--gpus` mandatory and ensure, that run will be executed on the right GPU.
 
+## Questions?
+
+Please have a look into the [FAQ](docu/FAQ.md). Also feel free to open an issue to discuss additional questions. Your question might make it into the FAQ as wel ;).
+
 # Citation
 
 If you benefit from this code, please cite our paper:

docu/FAQ.md

@@ -0,0 +1,28 @@
+# FAQ
+
+### How can I use KittiSeg on my own data?
+
+Have a look at [inputs.md](inputs.md) and this [issue](https://github.com/MarvinTeichmann/KittiSeg/issues/8). Feel free to open a further issue or comment on [issue 8](https://github.com/MarvinTeichmann/KittiSeg/issues/8) if your question is not covered so far. 
+
+Also, once you figured out how to make it work, feel free to add some lines of explanation to [inputs.md](inputs.md). As owner of this code, it is not easy get an idea of which conceptual aspects needs more explanation. 
+
+### I would like to train on RGB data, is this possible?
+
+Yes, since commit f7fdb24, all images will be converted to RGB upon load. Greyscale is those supported out-of-the box. This means, that even for greyscale data each pixel will be represented by a tripel. This is important when specifying the input format in [your hype file](../hypes/KittiSeg.json). Black will be stored as [0,0,0], white [255,255,255] and some light grey can be [200, 200, 200].
+
+### Can I use your code to train segmentation with more then two classes
+
+Yes, I had an earlier version run on Cityscapes data. This code is not compatible with the current TensorFlow and TensorVision version and I did not find the time to port it, yet.
+
+However making this run is not to much of an afford. You will need to adapt `_make_data_gen` in the [input_producer](../inputs/kitti_seg_input.py) to produce an `gt_image` tensor with more then two channels. 
+
+In addition, you will need to write new evaluation code. The current [evaluator file](../evals/kitti_evals.py) computes kitti scores which are only defined on binary segmentation problems. 
+
+Feel free to open a pull request if you find the time to implement those changes. I am also happy to help with any issues you might encounter.
+
+
+
+
+
+
+