New issue
Advanced search Search tips
Starred by 4 users
Status: New
Area: Infrastructure
Priority: Low
Type: Feature

Sign in to add a comment
Gold: perceptual diffs?
Reported by, May 1 2014 Back to list
Brian reports that he would like a better perceptual diff within rebaseline_server.

It should be fairly straightforward to add new perceptual diff columns; we just need to add new metric calculations to .

Brian, the following info (and anything else you want to add) would be helpful:
1. specific examples of image pairs where the existing perceptual diff let you down
2. pointers to alternate diffing algorithms that might be useful

Project Member Comment 1 by, May 1 2014
I haven't investigated 2). Ravi do you have a link to the cluster telemetry mesa vs. sw-raster results? That should serve for 1).
Comment 3 by, May 2 2014
Brian, if you can please pull a demonstrative image pair out of those results and attach it to this bug, that would be quite helpful.
Project Member Comment 4 by, May 2 2014
So the main problem I see is that image-heavy sites wind up being slightly different almost everywhere. The attached examples have the second highest pdiff score in the telemetry results. (The second image will be in a follow on comment b/c of the upload size cap).
9.7 MB View Download
Project Member Comment 5 by, May 2 2014
9.8 MB View Download
Comment 6 by, Aug 18 2014
Labels: Epoger-Untriaged
Comment 7 by, Aug 18 2014
Labels: -Epoger-Untriaged Epoger-RebaselineServerUI
Comment 8 by, Aug 21 2014
assigning RebaselineServerUI bugs over to Stephan; see
Project Member Comment 9 by, Oct 6 2014
Labels: Area-Infrastructure
Project Member Comment 10 by, Jul 6 2015
Labels: -Component-RebaselineServer Component-Gold
Project Member Comment 11 by, Jul 17 2015
Summary: Gold: perceptual diffs? (was: rebaseline_server: add more perceptual diff metrics)
As I understand it, Gold does not do *any* perceptual diffs, does it?
Project Member Comment 12 by, Jul 17 2015
Sign in to add a comment