I came in today so that I might be able to solve the mystery as to how to parse the XML files. I got most of the classes written for C++, but I still can't get the sample code from the Xerces page to compile. I am going to ask Viet if it is a problem with my Makefile tomorrow, but it doesn't seem to be a linking error which would have made more sense to me given what is going on in the code. If he can't help, I think I will have to email Konrad again and see if he can help (maybe he didn't do a full build and that is the problem).

I actually got through to Yi today and we went over the different tags in the different kinds of files! Yay! That was really exciting. Now I completely understand what the datastructures need to look like, and I think I know what attributes Sal is likely to want to use and which attributes probably won't be as useful to him. Also, I brought up the issues that I have been having with Xerces to Yi, but he couldn't offer too much help on that front. He suggested that if that parser totally doesn't work (though, he too, thought that it should since it is fairly widely used), that I might consider just parsing it in Java and outputting text files that are then read by C++ and use my own little delimiters to indicate the different tags. In a sense, this means parsing the data twice. We'll see what I end up needing to do.

Konrad got the code to compile! I should be able to complete the parsing now! Yay!

We had a meeting today, after which I met up with Sal for a bit. It was at that point that I realized that Konrad fixed everything. Also, Sal gave me the reference to a multiclassification paper that I really need to check out for the next step in our project.

To do:

1. Get the parser working and everything into the C++ datastructure.
2. Provide statistics for the structure operator.
3. Read multiclass paper.
4. Formulate rigorous idea for multiclassification to Sal and make sure I can implement it.

Realized that yesterday's compiling, and, in particular, my Makefile, weren't quite up to snuff. Fixed that today, such that I can successfully compile more than one C++ file:)

Tried to get all the parser stuff to work again today. Unsuccessful for quite some time. I was not sure how to convert XMLCh* to char*, and that was holding up progress. One of my officemates took a look at I was doing, (I think it might have been due to my vocal frustration and his general kindness), and showed me how to do it! Yay! Now I can make progress!:)

We had lunch with the group of DMP girls today. That went alright. We met some girl, (I think Sara?), who is working in the CS dept over the summer. I think she mentioned putting Arati on a help list for her compiling problems...that was nice of her. We also saw Morgan, whose number we forgot to get for this past weekend. It seems Morgan would like to go on a DMP outing this weekend--could definitely be cool:)

I worked from home today because I wasn't feeling very well. I actually got a lot of the code working. I am still not totally done with the parsing, but all of the classes that I wrote for the data structures are working out quite well, and I am able to parse some of the file. It is just a matter of finishing it up, from what I can tell, at this point.

I got behind in my journal, so I am now going back and filling in what has been going on.

I am finishing up the parsing and data structures in C++. I am focusing first on getting the color regions processed, but I don't think Jenny has gotten those files generated yet, so I am just using example XML files from Yi's machine to test the parsing. The focus right now is really the structure regions anyway, but the texture and color seem to have more overlap than with structure at this point. I am still figuring out a few things about the parser, but all is basically going smoothly at this point.

Today we had a meeting with Linda and the rest of the 2D group to talk about what we wanted to see at this point. We are trying to evaluate the structure features. Sal would like to run some experiments on "manmade" vs. "nonmanmade" objects. The distinction isn't quite perfect, but approximate. The basic gist is that we want to see whether the structure features are good about detecting nature vs. things like buildings, possibly cars, etc.

I am not quite done with parsing the structure aspect of the XML files yet, but I am getting closer. The color and texture seem basically done. Ran into a couple of problems when I saw that the way in which Color is stored in the structure features was different from the way in which it was stored in the other two, but was able to convert and maintain a consistent representation on the C++ side. Initially I think I was told that he was using hue and luminescence for color features in some of the files, but that doesn't really seem to be the case. Therefore, I have kept everything in RGB format.