Week 5: Continued Work on Research Project!
I spent most of Monday continuing my work on the models. They are taking longer than I expected. Back when I went to ZBrush classese at CDA, I could usually produce a mostly-done sculpt in 24 hours or so. Well, I've been working on these Peer Tutor agents for far more than 24 hours, so I wonder what's going on...I think it might be because I"m trying to achieve a certain style/look for the agents. Usually, I don't need to think about constraints or parameters when I model - I just go for as realistic a look as I can. I'm usually also making some sort of creature or a deformed humanoid. My hypothesis is that the constraints I've been given - the models must look like preteens, must not look too realistic - are making the modeling process more deliberate and somewhat more time-consuming. But I am definitely making progress, for which I'm grateful! I'm still not completely satisfied with the models, but I'm really hoping to finalize them this week. Yeah, if we're going by the timeline I proposed earlier, I am definitely a little behind schedule; however, I tend to be overly ambitious in setting up timelines. Moreover, my timeline allows for delays in deadlines - so I think all will be fine.
What sucks about my brain is that it tricks me into liking the art/writing/project I'm working on at the present moment - no matter how crappy the art/writing/project actually is. Actually, that's not quite accurate: instead of just making me like whatever creative "thing" I'm making, my brain tricks me into thinking it's somehow good! Example of my brain's devious nature: my models. I've been working on them for about a week now (not counting the intial design period from last week). I leave the lab every single day thinking that my models look pretty good and that they're probably on the brink of completion. Looks like I can relax tonight, my devious brain will think. The next morning, BAM! I open up my modeling files and find that my models look nowhere as good as I thought they did. Usually, there's some deformity going on in the face - skewed lips, wonky eyes, proportions that just don't work. I illustrate my brain's deviousness through the series of images below. The rightmost pictures show the modelsI had at the last intern update (which occurs every Thursday). Can you believe that, at the time of the meeting, I was actually quite pleased with them? Oh boy...those wonky eyes. The middle pictures show the models I sent Dave on Monday night. Again, at the time of the sending, I thought they looked quite good. Finally, the leftmost pictures show the models I worked on for all of Tuesday. I think there's a fairly significant difference between all variations of the models; but at every step of the way, I was convinced my models looked good and were on the brink of completion! Well, let's hope that the final models I produce are models that I'm satisfied with at any time of the day - not just when I'm making them! Unfortunately, the final models will almost certainly not look like the rightmost pictures (the ones I am most satisfied with). As Dave has informed me, there tends to be an "uncanny valley" effect when agents look too realistic (think Polar Express). Because the realism of their movements don't match the realism of their appearance, viewers experience dissonance and can't "buy" into the realism of the agent as a whole. Now, I didn't think my models looked that realistic, but I can see what Dave means. It's better to be safe than sorry. So the plan for tomorrow is to make several rough variations for a style continuum. My current models will represent one end of the continuum, the "very realistic" end. At the other end will be "very stylizied" versions of the same models. I'll probably have three or four intermediate stages between the two ends. Should be fun!
I spent all of Wednesday working on the aforementioned continuum. I had fun with this task because it involved playing around with the models - stretching their faces one way, replacing their realistic eyes with beady eyes, etc. I ultimately ended up with three other models for each gender besides the realistic models I had created earlier. That is, my continuum had four styles, which I will call here Style 1, Style 2, Style 3, and Style 4. Style 1 is a very cartoon-inspired. The models were intended to look a little dough-like and plastic with black, beady "button" eyes. I don't imagine we'll end up choosing this style, but I think it represents a good extreme for the other end of our continuum. Style 2 is still semi-cartoonish (think Wallace and Grommit). Less plastic, more dough-like. I replaced the bead eyes with semi-realistic eyes (i.e. the pupils and whites of the eyes are defined). I don't completely love this style, but it's definitely more workeable than Style 1. Style 3 is essentially a toned down version of Style 4 (the realistic model). I toned down a few of the features, including smoothing of the nose and "tucking in" the eyelids. I personally like this style the most of the ones I've listed so far, but I can see why it might not work (too similar to Style 4). I don't think I need to talk too much about the last style, since it's just the realistic style that I've been using the past week or two. Overall, I think I have a decent set of models to show everyone, and it will help a lot to get feedback from Dave and other members of the lab. Dave and Evelyn also brought up the possibility of me doing more CS-intensive tasks. The idea is, of course, appealing, but I'm really quite happy to help the lab in any way I can, even if it's not necessarily coding. We will be talking more about these potential side-projects tomorrow.
I spent most of the morning coming up with another model style for the continuum. I realized that Dave had liked Option 3 from the original model sheet (the one with tons of pictures depicting various 3D characters). None of my current models look like that, so I tried altering my Style 2 model to look like that. Later on in the morning, Evelyn and I met to discuss the CS side projects I mentioned in Wednesday's blog post. She had asked me to brainstorm a few ideas the night before. I'm not sure if I've mentioned this, but I've been trying to expand my skillset this summer by learning HTML, CSS, and PHP. HTML and PHP are surprisingly intuitive, and I've been happy experimenting with them. I thought I might be able to apply my newly learned skills to the benefit of the ArticuLab, since I noticed that the Rapport project doesn't have its own page on the ArticuLab website. This was an idea I mentioned to Evelyn. Other ideas we mentioned included setting up a better data collection system for Dave's user study. This user study is somewhat related to the Peer Tutor agent - in it, 7th to 10th graders will be teaching each other math. I expect we'll be using the rapport data we gather from these sessions for the creation of the peer tutor agent.) Recruitment of participants for the study involves a lot of data - their names, background information, availability - so I can see how a solid data organization system would benefit the lab greatly. We came up with a brief list of ideas that we decided to pose to Justine later on at the lab meeting. The lab meeting was helpful for clarifying a few things. First, Justine preferred Style 2, which means that a few additional days will need to be spent modifying my models - like Dave, Justine brought up the "uncanny valley" effect. A slight divergence here: I have to mention that I'm strangely conflicted regarding the models. On one hand, I'm still rather fond of my original "realistic" models; it'll be hard letting go! On the other hand, I'm oddly flattered that people think they look real enough to potentially have the uncanny valley effect. I think I'll keep the realistic models for future works. Back to the pertinent topic. In addition to the models, Justine gave us the OK for the side projects, which means that I can help Dave with additional tasks (which at this point I am still unclear on). Third (and this point serves more to introduce than clarify), Justine is giving a talk for Solve for X and may need a few images, which I am very happy to provide. During this week, I've also started to think that I'd like to help out more with Dave's user study. I can see how hectic it gets in the lab when there's a study under way - Cameron and Shannon stayed at the lab until 11 PM one night! While I enjoy making my models, I feel a little self-involved when I work on them. Angie and Marissa, the other two Rapport interns, seem to have their hands full with tons of different tasks - if I can, I'd certainly like to help them out. This was, accordingly, a sentiment I brought up to Dave after the lab meeting. He agreed, meaning that I'll probably be more involved with aspects of the study in the upcoming weeks. Another important development in my REU experience. :) Justine gave me a few outlines for the images she wanted, so I sketched some very, very rough thumbnails later that night while watching Futurama. I'll include them in the slideshow for this week's blogpost, so watch out for those. As a side note, remember what I said about my brain tricking me into liking my own work? Well, it's an observation that also applies to blog-writing. I think blogging is always kind of an awkward process for me, so it's not that my brain ever tricks me into liking the blogposts I write. But it always seems like my blog posts are that much more awkward when I go back and read them! Blogging shouldn't consist of purple prose, I know, but I suppose I'm uncomfortable with the informality of it all. I don't really "polish" these blogposts as I do with my more formal writing pieces; and if anything, I enjoy the stream-of-consciousness method of writing. But wow do they sound awkward when I go back and read them!
I spent most of Friday helping out with various side tasks. This involved making a flyer that will be used to recruit participants for Dave's study. I also spent some time looking at examples of volunteer recruitment posts on craiglist. When I wasn't working on mini-tasks for Dave's study, I spent my time modifying my 3D models to fit Justine's preferred style. Dave also mentioned that he wanted some sort of "share" bar available on his participant recruitment form (so that participants can email the form to friends), so I'll be looking into that this weekend. Other than that, I don't have too much new things to say about my work, except that everything is going well and I'm still having fun!