My work.
It seems that my work with the bee counting program has come to a halt. After thinking about it and talking to Michael, the person who worked with this program previously, there isn't much I can do to improve the program's chances of accurately counting the bees, unless we have access to an infrared camera...
... which then, will result in such an image. This, I think, would be an ideal solution to track bees precisely, because only the bees emit red heat, while the surroundings are cold blue. This kind of stark contrast is precisely what the program needs.
Then, after messing with the color levels, such an image could be produced. Hold on, there's one more step...
After messing with the contrast of the picture, you get this. You can invert the colors to get black on white. Right now, this is the best way for a machine to count bees. As you can see, there are still problems when the bees are too close to each other, as two bee blobs form into one ugly-looking abomination. Presently, this is not a problem that a machine can solve. Perhaps there exists a much more sophisticated way to detect bee, but I don't think I'll be developing that algorithm anytime soon, hah.
Michael said he's played with the bee images a bunch of times to try to make counting better, with no results. If Michael, who has more experiences and resources than I, can't do it, how can I? I know this is defeatist thinking, but if I spend a lot of time trying to improve the accuracy without any results, that would be frustrating. Then again, if I don't even try, I won't find out. Hmm...
In any event, I took a two-week break from the counting program and did some work related to Allison's project. There exists some Ruby scripts that loads images from a computer to a website, and creates thumbnails for them. My job was to modify these scripts to make them work for audio files. Audio data has a whole different set of metadata (data about data) than image files. Regarding image files and audio files, the former has things like width, height, exif version, color space, orientation, and the latter has things such as length, bitrate, sampling rate, number of channels, etc.
I had to work with a completely different set of resources to extract such information. I found out the hard way that it is very difficult to work with Ruby in a Windows environment. Much of the resources and instructions I found online were tailored for Linux. Sooo, I decided to to install Linux on my computer. At first, I installed Linux so that I could dual boot it (which means I can choose Windows or Linux when I turn my computer on). Then I changed my mind and just used Virtual Box, which is software that can emulate different operating systems.
BUT installing Linux on my laptop in the first place caused so many problems. Long story short, sometimes my computer thinks that my copy of Windows 7 is not genuine at it refuses to start up when I wake it up from sleep. Then after I force-restart it prepares to configure Windows, as if I had just installed a new copy. It gets stuck in that screen and I have to force-restart it. Sometimes I can fix this problem, and other times I can't access my laptop. This problem is reoccurring and extremely frustrating, and I won't be able to fix it permanently until I go back to the US and reinstall Windows using my installation disc.
Anyway, that's my problem and I'll deal with it.
I'm very close to completing my work with Ruby, and I am getting ready to work with the bee counting program some more. If I can't improve the counting accuracy for the photos in their current state, maybe I can create features that will work for infrared photos, in case infrared cameras get cheaper in the future and will be able to be used for regular fieldwork. That's the most I can do, it seems.
Well, back to work. Peace out.
In any event, I took a two-week break from the counting program and did some work related to Allison's project. There exists some Ruby scripts that loads images from a computer to a website, and creates thumbnails for them. My job was to modify these scripts to make them work for audio files. Audio data has a whole different set of metadata (data about data) than image files. Regarding image files and audio files, the former has things like width, height, exif version, color space, orientation, and the latter has things such as length, bitrate, sampling rate, number of channels, etc.
I had to work with a completely different set of resources to extract such information. I found out the hard way that it is very difficult to work with Ruby in a Windows environment. Much of the resources and instructions I found online were tailored for Linux. Sooo, I decided to to install Linux on my computer. At first, I installed Linux so that I could dual boot it (which means I can choose Windows or Linux when I turn my computer on). Then I changed my mind and just used Virtual Box, which is software that can emulate different operating systems.
BUT installing Linux on my laptop in the first place caused so many problems. Long story short, sometimes my computer thinks that my copy of Windows 7 is not genuine at it refuses to start up when I wake it up from sleep. Then after I force-restart it prepares to configure Windows, as if I had just installed a new copy. It gets stuck in that screen and I have to force-restart it. Sometimes I can fix this problem, and other times I can't access my laptop. This problem is reoccurring and extremely frustrating, and I won't be able to fix it permanently until I go back to the US and reinstall Windows using my installation disc.
Anyway, that's my problem and I'll deal with it.
I'm very close to completing my work with Ruby, and I am getting ready to work with the bee counting program some more. If I can't improve the counting accuracy for the photos in their current state, maybe I can create features that will work for infrared photos, in case infrared cameras get cheaper in the future and will be able to be used for regular fieldwork. That's the most I can do, it seems.
Well, back to work. Peace out.
No comments:
Post a Comment