10 Replies Latest reply on Jul 17, 2012 3:25 PM by fafnir88

    GSoC - Automated Visual Verification

    fafnir88 Newbie

      Hi Arquillian community,


      My name is Jakub Dunia and I am participating in Google Summer of Code this year, with Automated Visual Verification project.



      For all interested, my proposal :




      https://community.jboss.org/thread/197801 - another thread started by my mentor Lukáš Fryč, with more informations.


      Repository : https://github.com/cube88/arquillian-rusheye


      Summary : My goal is to build a review tool for result set generated by RushEye and improve algorithms used for visual verification (mainly speed them up and allow using many algorithms on one page in different regions).



      Well, for now that's it   I am still reading tutorials and documentation, so there is not much to report.


      I will report about my progress at least once a week here.


      I look forward to any comments and suggestions about the project. Since I have no prior experience with RushEye, all hints from experienced users are welcome.




      Jakub Dunia

        • 1. Re: GSoC - Automated Visual Verification
          Dan Allen Master

          Welcome and congratulations, Jakub!


          We plan to do a hangout (Google or alternative) soon so you touch base with the core Arquillian devs. We'll be in touch.

          • 2. Re: GSoC - Automated Visual Verification
            Lukáš Fryč Master

            Hi Jakub,


            congratulations to acceptance and welcome to Arquillian!


            I'm looking forward to hear from you about findings.


            If you need any immediate help to dive into the project quickly, you can find me on

            irc.freenode.net / #jbosstesting

            • 3. Re: GSoC - Automated Visual Verification
              fafnir88 Newbie

              Hi again,


              A little update about my progress. All the small stuff done.


              Repository forked : https://github.com/cube88/arquillian-rusheye


              Test build, test run, simple comparisons done. Since there are no more preparations to do, I started working on Manager (first part of the project).

              • 4. Re: GSoC - Automated Visual Verification
                Dan Allen Master

                Great! Thanks for the update!

                • 5. Re: GSoC - Automated Visual Verification
                  fafnir88 Newbie

                  Hello everyone,


                  My first 'real' update... Although starting earlier didn't quite work out because of finishing my M.Sc., last week I started working normally on the project's frist part - Manager. Current result look like that :



                  What is done :

                  • Simple project management - we can create project from directories, parse them and aggregate pictures into tree of test cases.
                  • Simple version of DoubleView - first view where we can compare images.
                  • SingleView - where we can switch between patterns and results (something like layers in gimp), color differences etc. [that's the full idea, now it only show diff image based on pattern and sample - rush-eye core code reused for this].



                  Generally, strategy for the first part of this code period is to make all the user interface stuff, so that anyone can try it get the idea about usage.

                  • 6. Re: GSoC - Automated Visual Verification
                    Lukáš Fryč Master

                    Great job, Jakub!


                    I will look into setting up additional test suites, so we can try on different real data than you are using as base for development.

                    • 7. Re: GSoC - Automated Visual Verification
                      fafnir88 Newbie

                      And the next update is here.


                      What was done last week :


                      Double View with synchronized scroll (we change position of one image and the second follows).

                      Single View with sample/pattern/diff image, zoom in/out, and focus button, that sets display to part of image with errors.

                      Generating Visual Suite descriptor through menu.

                      Marking tests with buttons as positive and negative (ResultConclusion field in TestCase where I store data, not in actual result descriptor).



                      On this screen we can see generated xml from directory, and in the right menu we have buttons and labels, where we can see/set status of selected test.

                      • 8. Re: GSoC - Automated Visual Verification
                        fafnir88 Newbie

                        And another one...


                        Still making usefull gui:


                        - Drag and drop for moving image.

                        - custom icons in  tests tree.

                        - filters in tree (only negative/positive/not_tested).

                        - menu under each image view.

                        - Integration of parse function from RushEye.


                        - better image management

                        • 9. Re: GSoC - Automated Visual Verification
                          fafnir88 Newbie

                          And update again (looks like I missed one considering one update per week pattern).


                          Anyway, project is progressing fine :


                          We can now perform all operations through manager.

                          - Generating test suite

                          - Running parse of this suite

                          - Loading suite and then see/compare test cases.


                          Small bugs like problem with unpredictable drag and drop fixed.


                          Still need to do crucial part of inserting changes made by manager into result suite.

                          And after that generally make gui more simple and user friendly.

                          • 10. Re: GSoC - Automated Visual Verification
                            fafnir88 Newbie

                            Midterm over, looks like I'm still in.


                            Manager is in quite advanced state, on github https://github.com/cube88/arquillian-rusheye there is manual with all current working stuff.


                            To sum up:


                            Basic stuff is done and working, user can generate suite, load it, compare images and set his own results for patterns. Then he can run parse for suite and do normal parse(that takes into consideration results user set).  User can also set result xml file and change already written results. That's basic stuff, still more to do, and to do it more simple.


                            Some changes in schedule, still working on manager and mask creator as part of it(mask creator was scheduled last originally).