- Visualization script takes
.t7
file as input which was generated usingvideo2Tensor
script. - Set
channel[1]
to 1 if its a gray-scale image and to 3 if its an RGB image. - Use
nrow
option to control # of images to show in one row.
Sample command to run the script:
qlua visualize.lua --input /dataset.t7 --dmodel /media/HDD1/Models/predNet/ --net 100 --nrow 5
This script shows a row of original images, a row of error images, a next-frame prediction and an hallucination (given a prime of a sequence length). You can run the script as:
qlua view_predictions.lua --input ../utils/testData.t7 --dmodel results/010 --net 20 --nrow 5 --model pred