I have successful completed the RGBD camera mount!
As you may recall this has taken me a long time to get to this point. First I had to find the correct 3D printing file, test it, correct it and re-print it. Once that was finally taken care of I had to track down the extra hardware that needed to go into it, namely some obscure screws and fasteners as well as a sliding plate mount for a DSLR camera. What I have in my greedy little hands now is an assembled mount that can hold both a Kinect and DSLR on a tripod securely. Progress!
Next I need a DSLR camera that supports live mode. Since I don't have $600 dollars to spare I've asked my friend Sheva if he would be so kind as to loan me his camera for a few hours while I do some tests. I'm still waiting on him to bring it to school, but I expect that I can get started with that tomorrow or early next week.
But what will I do with this you ask? Well I plan on taking the setup outside of the project room and recording some students doing whatever it is they do. From this footage I should get an idea of how well my final setup is working as well as the logistics of moving it all around after calibration. If all goes according to plan, then I have a few ideas for production videos, such as some CDM 3D digital life shots as well as a night time fire performance. The latter is a big question mark since normally you need lots of UV to get good footage. The lack of sunlight but addition of light from the fire may make for some compelling 3D images. Who knows!? Let's just hope security doesn't arrest me when I light my poi on fire. This should also give the back side of the CDM apartments something interesting to look at during dinner.
But that's not all! I've also been working with 123D Catch, continuing to try and nail down the process that will allow me to quickly create 3D models from a series of images. Here are some images of my most recent process:
I tried to be a clinical as possible with this shooting, even going so far as creating my own makeshift lazy suzan that allows me to incrementally rotate the object in a semi-controlled manner. What I ended up with was 64 photos that I thought would be great for 123D catch. So I sent off the files to the cloud and eagerly awaited a response from Autodesk.
The good news is that I got a model back from Autodesk and was able to start working with the stitching feature. This is great because I don't always get models back from 123D Catch, as they sometimes hang in processing forever.
The bad news is that it looked like this:
Yuck what is that? An amorphous blob is not what I was hoping for. After some consideration I came to the conclusion photographing a white object on a white piece of paper was probably not the best way to go about things. I suspect that contrast has a lot to do with the algorithms that Autodesk is using to create these models, and I missed the boat on that one. But no matter! I am using the desktop application which allows me to do some manual stitching of the photos.
This did not give me such good results. I am unsure if the program is just buggy as all get out or if my lack of contrast is compounding the problem. After 3 hours of trying to stitch photos, re-submitting the data and getting no noticeable improvement on the model I decided that I need to re-shoot (again) but this time with much more contrast. So that means either a black object or a black table on which a white object is shot.
So some progress and some set backs. I'm still trying to get caught up from my absence last week so I apologize for the lack of content. However assembling that camera mount was a big step in the right direction in my opinion. More to come next week! With actual videos! Promise.
No comments:
Post a Comment