There was a problem with the DIC sensor so I am unable to validate my data from the first sensor deployment. It will be deployed again tomorrow and hopefully that data will be able to be validated. I am still hopeful that I will be able to complete building sensor number 2 during my time here but that looks potentially problematic. I can't believe how close we are to the end, can I just have another 10 weeks here?

This week has gone by really really fast! Mostly I have spent it refining my Matlab code so that my data can be as accurate as possible. The noise on the baseline of the spectra is tricky to deal with when I have tiny peaks from propane, iso-butane, etc. Another complication is that because the methane is in the gas phase, rather than dissolved in oil as the Brewer lab has looked at before, it has more rotational/vibrational modes available. This creates a whole slew of little peaks that overlap with some of the other hydrocarbon peaks. Thankfully, I've been making improvements by calculating a more consistent baseline and by using peak height, rather than peak area, ratios to determine relative concentrations of the gases in my samples. I'll probably be taking a couple more standards samples next week, but I'm just about ready to finalize the code and crank through the data!

This week has been pretty productive. I am almost done comparing my manual data with the code data. I have successfully gone through the code data and essentially picked out the data only where jellies are identified and saved all the information in those frames. Knowing the frames I then could go through my manual data and collect corresponding points. The problem when comparing the jellies identified in the code and my manual identification was that the code gave centroid positions and I got the head and tentacles position, so I had to average my points so they could be compared to the centroid. I now have to go through that data and see if the centroid and average values are in-between some threshold and if they are then I can assume they are the same jelly! After this I will really be able to see the difference between the code and manual orientations of the jellies!

This week I revisited the ASP code for the Samples Database. The bug in which an error page appears whenever there's no images is still standing. I've been trying to fix the code so the error page wouldn't come up but to no success. I haven't revisited the Expedition Database as there are no longer any major bugs.
For the Django portion of my project, I mostly been making small edits for SampleDBLoader.py through the terminal.
My Progress:

https://github.com/livyliv83/stoqs/blob/samples_database_load/stoqs/loaders/SamplesDBLoader.py

Week 7:

This week I took a look at those stained samples from the cruise. Unfortunately, not much to see I've also been continuing progress on my projective reconstruction code. Big thanks to Prateek and Steve for helping me understand the deadly linear algebra + 3D geometry combo. I was able to eliminate the functions that I found online (I understand enough to make my own version!), and I'm getting results that at least look like the correct shape. I know that they're wrong, but I'm working on the triangulation and essential matrix calculations to improve them. Another issue is that my camera calibration was a huge approximation, so next week I'm going back to the SEM specifically for reconstruction images, so I can be careful to know where the cameras are in relation to each other and rotate about the center of the sphere. This is probably definitely what I should have been working on all along, but I didn't find the idea until halfway through the internship :/

I also learned this week how to make alignments and trees showing the relationships between gene/protein sequences, specifically the silicatein-like sequences from radiolarian transcriptomes. Pretty cool what kind of information that sort of analysis can give, and what misinformation if you don't look carefully!

Week 7

In week 6, I finished the regression plot with observational data.  I started working on converting all of my data into micromol/kg instead of ml/l (for whatever reason, people use almost 5 different units for dissolved oxygen concentration, but micromol/kg is the SI unit), however, I found that the density in the database is not reliable.  Thus, I should just use the density that I calculated for all of my analysis and figures, but I was under the assumption that the database values would be the most accurate and have used those densities in some places.  A mess.

On Monday and Tuesday, I made a regression plot with model output data.  It shows a different pattern than the observational data, but after talking with Francisco, I have some possible explanations and great ideas to move forward and test some of those explanations.  I am trying to finish the ROVCTD oxygen comparison this week, but I'm not sure if I can.  I talked to Marguarite and Jose about the way that they have done their oxygen comparisons, which was very useful.  Mostly I am limited because I do not know how to make the SQL queries, and am going through a UI which "dumbs it down" for me, but I can't figure out all of the functionality - really simple things I don't know how to do through the UI.

I downloaded a particle tracing software written in Python from GitHub which I will test and try to run next week.  If I can get it to run, I can test hypotheses about changes in ventilation in the North Pacific and changes in the strength of the California Current or Undercurrent, all of which are prime suspects in the changing dissolved oxygen mystery.

I'm really really excited with my project right now, because the code is finally reading through the data and I'm finally getting to the calculation part of the code so I can finally compare it to the WOA data so I can finally finally calculate a correction factor from the 2 data-sets that I can then apply it to the Argo data...finally B)

What an excellent week! I accomplished a lot more than last week and I am feeling pretty good about my project. I have started writing my paper and thinking about the presentation. I have to keep reminding myself there is still plenty of time to do everything! I am pretty excited because I will be using MATLAB this weekend and hopefully Brian's tutorial will help me figure out how to analyze the POC flux data from Station M.

It's been a slower week than usual. Most of my time has been focused on trying to perfect my abstract for submission to AGU by August 5th. I think it's almost there, but there are definitely still a few things that could still use some tidying up. I'm well beyond the 250 word limit suggested by most academics, but within the 2,000 character limit set by AGU. Perhaps ill use the longer one for the AGU and shorten for the paper?

Amy did a nice job tricking me into taking her spot in today's outreach program. I was given 5 minutes notice before having to sit on a panel, judging a hypothetical land use scenario given by Watsonville high school students. It was a lot of fun though those kids did terrific.

I really want to take some time this weekend to work on the introduction and methods of my paper. My discussion section isn't going to be as thorough as I hoped because I didn't get the OK from Charlie's collaborator to consider the depressions to be a major geohazard. I can still talk about the possibility, although its going to be a pretty weak case.

Week Seven:

Whoa, end of week seven this is getting close. I worked to filter all of my potential scale options no longer limiting my search to McMaster-Carr as their selection was fairly limited and a bit more pricey than the selection available on the vast internet. I presented the most fitting options to Brett and was able to get the order placed for two of these scales. Additionally, I feel the vehicle chock design and material selection has finally been narrowed to the point where we can start to order my materials and fabricate the chocks. I'm hoping to have all of the materials and the fabrication/assembly work complete by mid-to-late next week so I can begin to utilize the measurements and finalize my procedure. A lot to do, time is closing in!

I started the paper on Monday and spend most of my day writing the paper. I did some testing on the oscillation issue on the load that I designed. I worked on the presentation and have the outline done. I spend a few hours on drawing the system diagram of my project and including that in my paper. I also did some research on the oscillation problem and organized all the research I did in the past few weeks. So far, so good!

This week was really interesting as I finally introduced more classes( more than two species) into my classifier. I wanted the problem to be more challenging so that it can be a complex classifier rather than just a binary classifier. My classification system finally has four species namely Cuviar's beaked whale, Porpoise, fin whales and blue whales. The first two vocalize at high frequencies( 25Khz- 100 KHz) and the other two vocalize at low frequencies ( 20-50 Hz). I have used two classifiers( SVM's and Random Forest Classifiers) just to compare their function. I achieved pretty amazing results with both of them, just that random forest classifiers seem to work better than svm in the case when you remove some acoustic features( as Random Forest Classifier selects a fixed number of best features). I am planning to augment more data( that is create more data using the data we already have , which works pretty well and improves classifier performance). Lets see how it goes.

Week seven:

All this week I was analyzing my data and organizing my results. I would like to do more, but time is short ?. So, I was defining with Linda my goals within the internship. The rest we will continue later. Also, I started reading more literature about benthic communities and oxygen minimum zone to understand better my results and finish to write my paper.

And to close my short week, I was in an amazing tour for Western Flyer. It was excellent to know face to face the ROV (Dc Ricketts) that gave me all my data!!!

Hey gang,

Been under the weather this week but have still been able to get a lot done. I fixed a bug in my code written for matlab that was presenting the algorithm from finding a negative slope. I then went about translating that code into ruby, and am in the process of making the code actionable so that the rover can implement it.

In addition to the rover machine learning project I'm also putting together a summary of the physical environment from the rover's current meter.

This past week I have been trying to solve for the last two parameters I need to run my code: beta and alpha, the width and height of the image in radians. This has been very frustrating because I have already calculated these parameters twice already, was so excited when I got it, and then we realized there was something else we weren't considering so now I will have to take measurements again and recalculate these parameters. So my next step is to find a way to take pictures with the ROV camera, underwater with a 4.1 zoom, and the calibration board at a known perpendicular distance to the camera. Once I have this image and know that distance I can resolve for beta and alpha, run the code, find a way to relate this code to the measurements I took in the beginning of the summer, and finish my paper and presentation all within three weeks! I really want to get this all done!

Week Seven:

This week was spent collecting data for my experiment. I counted ~4500 fixed abalone embryos over two days, but I am now done with my data acquisition and moving onto Excel to create a graph comparing my three treatments.

The past week has been marked by a lot of drawing. While manual tracing is proving to be more effective than attempting to manipulate the videos in Matlab, it is tedious and time consuming. I am nearly done with the tracing and will begin applying different rotation and translation matrices to get an idea of variability in swimming cycles between organisms.