Re: Convolution based regtest comparison

On Sun, Jun 15, 2025 at 12:34 PM Luca Fascione <[email protected]> wrote:
> I do find your concept interesting though, I'm not sure what smoothing
> filter you're using, but I wonder if this is a case where traditionally
> "bad" filters would actually
> be advantageous here. Usually the "box" filter is considered the worse, on
> account of the substantial ringing it exhibits.

It is a box filter, on account of efficiency: since this is part of
the compile/debug cycle, it should be fast.

> Me, I had been musing if instead one could simply scan the EPS or PDF and
> put all objects into a quadtree,
>..

you can do very complex things, but

1) right now, it only has to be better than the whole-image MAE diff
2) as part of the development cycle, it has to be fast. That means it
has to be simple too.

if you want to create a quad tree of objects, it would be simpler to
revisit the original testing framework (which I removed), which
annotates Stencils with their bounding boxes and where they came from.

-- 
Han-Wen Nienhuys - [email protected] - http://www.xs4all.nl/~hanwen