Generate Image

[FYJ10]

Hi Remi!

I couldn't install docker because of some force majeure. I added tricks.py and otbtf.py in the python folder in otbtf to the code folder, and after a series of adjustments, I successfully ran train.py. (As shown below, the operation should be successful, right?)

I0721 21:39:35.738773 29916 train.py:281] training Epoch #0
I0721 21:39:40.855148 29916 train.py:291] training: one epoch done. Total FS stall: 0.00s
I0721 21:39:44.284425 29916 train.py:281] training Epoch #1
I0721 21:39:44.381165 29916 train.py:291] training: one epoch done. Total FS stall: 0.00s
I0721 21:39:47.372760 29916 train.py:281] training Epoch #2
I0721 21:39:47.468502 29916 train.py:291] training: one epoch done. Total FS stall: 0.00s
I0721 21:39:50.577277 29916 train.py:281] training Epoch #3
I0721 21:39:50.672024 29916 train.py:291] training: one epoch done. Total FS stall: 0.00s
I0721 21:39:53.689270 29916 train.py:281] training Epoch #4
I0721 21:39:53.787004 29916 train.py:291] training: one epoch done. Total FS stall: 0.00s

But I cannot use sr.py to generate fake HR images:
import otbApplication
ModuleNotFoundError: No module named'otbApplication'
I want to know if there are any other ways to generate fake HR images? Or is there a problem with which of my steps is incorrect? I really want to save these generated HR images!

[remicres]

Hi,
You could manually run the TensorflowModelServe application, from the terminal.
Take a look at the parameters values that the python API use in sr.py, and manually type the (long!) command line.
With OTB python API, it is easier because all you have to provide is the model and the image ;)