Re: regtest threshold problem
On 2024-04-28 03:34, Werner LEMBERG wrote: > Independent of improving the metrics, maybe we could improve the review workflow by adding dynamic control to the diff web page to expand or collapse the entries based on a variable threshold. Yes, this would be a nice feature – please submit an issue for that. https://gitlab.com/lilypond/lilypond/-/issues/6711 -- Dan
Re: regtest threshold problem
>> I'm working on a modification of `musicxml2ly` that will cause a >> change in the result as shown by the attached images. However, our >> regtest classifies this as 'below threshold' and doesn't display >> it. [...] > > Independent of improving the metrics, maybe we could improve the > review workflow by adding dynamic control to the diff web page to > expand or collapse the entries based on a variable threshold. Yes, this would be a nice feature – please submit an issue for that. Werner
Re: regtest threshold problem
On 2024-04-27 00:24, Werner LEMBERG wrote: I'm working on a modification of `musicxml2ly` that will cause a change in the result as shown by the attached images. However, our regtest classifies this as 'below threshold' and doesn't display it. Does anybody have an idea how to improve this? I'm not an expert in image analysis, and a quick search in the internet didn't help me in finding a good metrics. I think we would need something that checks whether a shape completely disappears and/or reappears in the images (contrary to shapes that are only slightly shifted and/or distorted), almost regardless of its size. Independent of improving the metrics, maybe we could improve the review workflow by adding dynamic control to the diff web page to expand or collapse the entries based on a variable threshold. -- Dan
regtest threshold problem
I'm working on a modification of `musicxml2ly` that will cause a change in the result as shown by the attached images. However, our regtest classifies this as 'below threshold' and doesn't display it. Does anybody have an idea how to improve this? I'm not an expert in image analysis, and a quick search in the internet didn't help me in finding a good metrics. I think we would need something that checks whether a shape completely disappears and/or reappears in the images (contrary to shapes that are only slightly shifted and/or distorted), almost regardless of its size. Werner