I spent the vast majority of this week labeling bees, and I was finally able to finish labeling all of the identifiable bees in the small swarm, ending up with 425 seperate bees.

Once this was done, I perfromed some analysis on the segmented swarm. I can’t put the majoroty of my more in-depth results here for confindentiality reasons, but here are a few basic outputs for the labeled swarm:

1) The distribution of bee sizes. After identifying each individual component in the swarm (note, some bees appear to be touching, so only 421 individual components were found), I wrote a basic function to calculate the sizes of each component and plot their distribtuion as a histogram with a fitted Gaussian curve. The average component size is about 671 voxels with a standard deviation of 167.

2) The centers of mass. After identifying each seperate component, I was able to calculate the center of mass for each bee and plot those centers as well as use them for a variety of other analysis methods.

3) The swarm centroid. Using the centers of mass, I began to try to find the center of the swarm based on the density of bees at different swarm depths. Based on this, I was able to determine how much of the base of the swarm to include when finding the centroid.

I also tried to begin training a new neural network using the small swarm as training data, but my notebook kept crashing due to exceeded RAM, so next week I’ll look more into how to make the data size maneagable.