Problem with the visualization of the FlowNet2 output

[sotomotocross]

I have installed the all the packages of the repo and i try to run the test of the flownet.
I run the generic image feeder with the basic data from the FlyingThings3D dataset and the i run the lmb_flownet flownet.launch file.
The NN runs properly but i have problem with the visualization of the outputs.
The rostopics are being published properly and the /flownet_vis topic also.
But the image of the output is distorted.

I cannot understand why this happens. I tried both the sintel and middlebury as viz_type.
However the visualization output is not even close on the expected output.

What do you think is the problem?
Did i do something wrong on the catkin_make and roslaunch procedure or something else.
This also happens when i use my own kind of data.
Also i want to note that i have succesfully used the flownet2 pytorch github repo from NVIDIA but it's an offline run and i used your repo in order to implement it in ROS/Gazebo.
I am looking forward for your reply and help.
Thank you in advance for the time.