User:TheLarry/Notebook/Larrys Notebook/2010/02/02

From OpenWetWare

Jump to: navigation, search
Tracking Main project page
Previous entry      Next entry

Automation

Automation keeps getting close to my level of satisfaction. That doesn't mean it is done, but most of the hard work would be. Next would be making it prettier and fixing any bugs that pop up. But right now it works pretty well.

The user can input a folder, and the program will find any directories with .pngs in it within that folder. So it will actually search through folders in folders to find any .pngs. So I think at this point you can run this over night and come in the morning to analyze a shitload of data. I also fixed that bug where it was tracking the same guy over and over again. What happened was a small piece of another microtubule was in the bounding box with the main guy. What i do now is find the object within the bounding box with the largest area and then black out every other object. (This was a problem in the tracking software as well, and if i put in this sub.v i can fix that as well. I gotta do that tomorrow.) Also I added in the dummy array that Koch suggested. before it was tracking bad microtubules over and over again since i threw out that data. but now i keep that data in a dummy array and don't track anything the software has tracked before. That is looking like it is working pretty well as well.

This .vi saves the data in a .dat file that I am sure Koch and I will wanna change the file extension since i just came up with something so i can get the program running.

Right now I am working on presenting the file paths in a single graph with each microtubule a different color. I figure the front and back of the microtubules will be the same color just one a bit lighter than the other. I got a prototype .vi running and it is doing this pretty well. Working with active plot and plot color property nodes is annoying. But this is doable.

So tomorrow. I wanna finish this visualization of the paths. This will be annoying more than just the graph property nodes but also i'll have to make an array of colors that is large but also will cycle through if there are more paths than array elements. Shouldn't be impossible just annoying. Also tomorrow i would like to add to the original tracking software (which i have also been calling the mother tracker or mother program) the ability to load a 2-d array data from these automations. This will allow the user to actually analyze the data that was compiled over night. Right now nothing can, but that shouldn't be too bad (knock on wood). A new event structure with the appropriate logic constants.

I don't know if I want to add this automation program into the mother program. Is it better to have it an independent program or have a new event that runs the automation? I am not sure. But I'll deal with that when everything is done.

Steve Koch 22:48, 2 February 2010 (EST): Holy Shit you got this done quickly! You were sooooo right and my pessimism was unwarranted--thanks for proving me wrong! My opinion is to have it as a standalone program, and then only slightly modify the mother program to be able to load from automatic results. But you're more familiar with it, so up to you.
Also, for absolute sure, you should feature this on your poster in Biophys. It could definitely be very useful to dozens of groups out there (perhaps more if extend outside the field of people we'd interact with: e.g. people who track snakes & worms). The big challenge is getting people to (a) recognize that it's good, and (b) be willing to invest time to get it to work. In terms of (b), if it's a serious user, I think we should be prepared for you to spend a couple days with the new user getting things to work for them. Once one or two other people used it, I think it'd make others more likely to put in the investment of time. Another idea is that someone could send you their series of PNGs and you could send them back the results in 2 days to blow their mind. I am itching to brag about your program. Can you make just a quick cam studio (not even with audio maybe) that we can post on youtube / friendfeed? I'd understand, though, if you want to wait some more till it's even more perfect.

TheLarry 23:23, 2 February 2010 (EST): I can cam studio it Thursday or Friday. Yeah I can pretty proud with how the program turned out. I could put it on the poster but how much should it take up?

Steve Koch 23:29, 2 February 2010 (EST): Whatever is appropriate. Don't worry that your abstract focused on the theory. I think you should have the modeling and the tracking on your poster, and whatever mix will present the messages you want to present the best. Try to imagine yourself as the poster viewer and what you'd want to see. You can also coordinate with Andy as to how much tracking he'd want on his poster versus yours. The two posters can have overlap, but can also be used if space runs out on one or the other also. And definitely you should be wicked proud.


Personal tools