Thanks for such detailed instructions.
I am trying to replicate the steps in the paper. The authors have trained a small MLP network (5-32-32-32-2). I guess you also have collected data for this network and probably trained it.
Would it be possible to get the script for model training? Did you concatenate RGBXYGxGy values of each image into one huge CSV file?
If I am not mistaken, GelSight images are 320x240, so we should get 320x240 row and 5 column array for each image, right? How many images were enough for the training in your case?
Thanks in advance!