Hi Rupert, I have read the documentation of Stanbol Enhancement Engines and also about Enhacement Structure. I am interested in two issues actually which are Stanbol-1006[1] and Stanbol-1007[2]. I have read the previous conversation[3] about the face-detection. >From what I've understood since the contentItem has getblob() method to get the main content which can be video/images/audio file and also clearly we don't have any problem of knowing the type of content by getting the mime type, now I need to create an enhancement engine which will be a part of enhancement chain and it will be using OpenIMAJ/CMUSphinx and then result(detected faces in images/written text of audio file content) will be stored in the contentItem and will be the final result. Regarding the face detection in video thing, one way to reduce the no of frames would be to check the difference between the consecutive images and if there is a considerable difference then we need to apply face detection to limited number of images only. I have tried OpenIMAJ and CMUSphinx demos. I have decent experience in Java. Could you please guide me how should I start writing my proposal. My first priority is Stanbol-1007[2](Speech to Text Enhancement Engine) and then Stanbol-1006[1].
[1] https://issues.apache.org/jira/browse/STANBOL-1006?filter=12326260&jql=labels%20%3D%20gsoc2014%20AND%20text%20~%20%22stanbol%22 [2] https://issues.apache.org/jira/browse/STANBOL-1007?filter=12326260&jql=labels%20%3D%20gsoc2014%20AND%20text%20~%20%22stanbol%22 [3] http://markmail.org/message/3vwdsjuqdamlvuxr#query:+page:1+mid:3vwdsjuqdamlvuxr+state:results On Thu, Mar 13, 2014 at 4:44 PM, Rafa Haro <rh...@apache.org> wrote: > Hi Rohan, > > El 13/03/14 09:54, Rohan Jain escribió: > >> Hi, >> I am a 4th Dual Degree(B.tech+M.tech) Computer Science student of IIT >> Kharagpur. >> I am very much interested to be a part of Apache-Stanbol Development in >> GSoC 2014.* I was going through the ideas section and I think I can >> contribute to "**Multimedia Enhancements for Stanbol" idea as I have >> >> already done a lot of work in 2nd goal of the Project-Idea >> <https://issues.apache.org/jira/browse/STANBOL-1003? >> filter=12326260&jql=labels%20%3D%20gsoc2014%20AND%20text%20~ >> %20%22face%20detection%22> >> which >> is to *"Develop Enhancement Engines for Multimedia Content"*.* >> >> >> In the open soft event our team developed a software for windows which can >> do around 12-15 commands by listening to the speech of user like opening a >> file,playing music,video etc, shutting down computer. In that I >> *implemented >> the speech recognition part* which converted it to text and it was done in >> >> C# where I used Speech Recognition >> Library<http://msdn.microsoft.com/en-us/library/system. >> speech.recognition(v=vs.110).aspx>. >> In my Image processing course I worked on two projects *one was of face >> >> detection where around 70 sample faces were used to train it and then it >> gave 80% accuracy*, it was done in Matlab. Other project was to *separate >> >> the stages of boiling milk from a video of boiling milk like warming to >> onset of boiling to onset of spillover to spillover of milk* which was >> also >> >> done in Matlab. So for the 2nd goal I have a very good experience and I >> strongly believe that I can contribute a lot to this. >> > As Rupert has just pointed out, those are quite concrete use cases that > for sure would require to train custom models. Note that there is also an > issue labelled for gsoc2014 about Topic Classification that could involve > the development of an API within within Stanbol for training Content > Classification models. That API could be adapted/extended for supporting > any kind of training job to produce models that could be used later in > enhancement engines. Just an idea :-) > >> >> *So please guide me on how I should proceed towards my proposal for GSoC >> and for this particular task what should I start doing?* >> >> > -- Regards, Rohan Jain 4th Year Dual Degree Student Dept Computer Science IIT Kharagpur Kharagpur,West Bengal India Contact - +917797256872
