[Suggestion] Alternatives to hog for faster runtime

[Seon82]

The new version of the program runs significantly slower than the previous version (with the speedup provided by #18). A bit of profiling reveals the hog method as the main culprit.
I've implemented a few alternative algorithms I thought might show similar results, and benchmarked the time necessary to do a fit_transform using these different methods. I feel like the 2x2 sobel could be a pretty solid alternative to hog, what do you think?

(All benchmarks run with dither="naive", palette=6 and boost=True)

There's a small mistake in the labels, 3x3 sobel is in fact 2x2 and vice-versa

Factor 6

Factor 10

Factor 6

Factor 3

Factor 10

Factor 6

Factor 10

Factor 6

[sedthh]

That's a really good idea. I am happy with the top row so it definitely worths a try!
Could you make examples with 3x3 sobel as well, and images where the edge detector is called multiple times (similarly to depth > 1)?

Thank you in advance!

[Seon82]

Just noticed a mistake in the labels for the images above, 2x2 sobel is in fact 3x3 and vice-versa. But there are already some 3x3 sobel examples above.
Will look into the different depths asap!

[Seon82]

Depth 2

Depth 3

[sedthh]

I love it! Could you maybe add both 2x2 and 3x3 as choices to be set s hyperparameters with Pyx() with 3x3 as default? The 3x3 looks better but 2x2 is just so much faster, it will worth the tradeoff for rendering animations later!
Pyx(sobel=3)

This is great!