Run model 2 with tensorflow for watershed data (from 2017/02/23 ~ )

[tomowind]

Running from 02/23 12AM. Running from sangmin-tf-ontopof-20170222 branch.

Here's the config...

[config]
img_rows=512
img_cols=512
model_id=2
batch_size=2
nb_epoch=500
standalone_loss_history_file=standalone_loss_history_file

[tomowind]

Result (02/26):

• 45 epochs.
• Loss not going down.
• Ran more than 3 days.
• Saved weights and losses.

Next:

• Load theano's weight? (try this: http://stackoverflow.com/questions/42193436/keras-convert-pretrained-weights-between-theano-and-tensorflow)
• Use tensorflow's weight to theano

[tomowind]

Add results to code in results directory in the following form:

results/model_2_tensorflow.md

- configuration: (write this if you modified the source to run the model)
- code: which branch, which commit
- result: 
  - loss: something
  - ...
- weight file: saved in sangmin's machine