Hypothesis: Training a model with an animation transformation from input to output, for all train pairs. This may be a good clue how to solve tasks. Unlike not having any info about steps to take to go from input to output.
Currently it's only the test pairs that are being recorded.
It can be interesting seeing how a human convert the train pairs from an input image to an output image. So the same rules gets applied.
Solution
Proof of concept
All the output images for train pairs will have to be editable. To begin with this will only have to be available in an expert mode. Later it may be made available for everyone, but I doubt it.
Problem
Hypothesis: Training a model with an animation transformation from input to output, for all
trainpairs. This may be a good clue how to solve tasks. Unlike not having any info about steps to take to go frominputtooutput.Currently it's only the
testpairs that are being recorded.It can be interesting seeing how a human convert the
trainpairs from an input image to an output image. So the same rules gets applied.Solution
Proof of concept
All the output images for
trainpairs will have to be editable. To begin with this will only have to be available in an expert mode. Later it may be made available for everyone, but I doubt it.