18 May 2018

The Latest

Well, Despite my best efforts, I have not been able to keep up with my blog as the year has worn on. I can’t say I’m surprised. Life is pretty busy now with a little one. We’ve actually been out of town a couple times in the past month or so with various events. On top of all that, I started my new gig on May 1st. The new job isn’t taxing yet, as I’m just plowing through all of the training material that I’ve been given. I keep hearing how everyone is super busy, but despite that, they don’t want to overload me with projects before I get through the training material. Quite the departure from my last role, where they basically threw me to the wolves on day 2. I’d say I’m getting through the material quickly, despite it being incredibly dry (engineering specs and online learning modules, wooo). I’m sure I’ll be getting antsy and want to do some real work for the team. In the meantime, though, I’m happy to have a little decompression time.

Finding Time

With all of the various things in life that require my time (work, child, hobbies, house chores, learning & other projects), I often get frustrated that I “don’t have enough time to get anything done”. Balancing the demands of work, child, and general housekeeping is tough enough, so really I was just frustrated that I wasn’t getting time to work on personal projects. One partial solution to that is that my wife offered to pick up our little one once or twice a week, when she can. That gives me a chance to get home a little earlier and with no child to watch, I can focus on a project, class, or whatever. That was a good start, but I still felt like I needed more time. I decided to try the novel approach of getting off the couch, putting the smart phone down, and actually do something after the little one goes to bed. On top of that, the “life hack” that has made this approach work so well is to focus on one thing per evening. It’s a little compartmentalized, but that seems to be the best way to focus my energy and actually make progress. It all started with “work” Wednesday, where I would work on one of my learning projects (data science, or the deep learning class, the Flint project with my brother). From there, I decided to assign a different thing to each week night. Monday is cleaning day, where I tackle a chore that I didn’t get to over the weekend. Thursday is music night, where I practice with my music gear. Tuesday is a flex day, that can be used for getting ahead on any of those subjects. I have found that focusing my energy in this way has also prevented me from dozing off on the couch watching tv, or staring at my phone. I actually have to tell myself to go to bed, to ensure a decent night’s sleep! I’m sure I’m not the first parent to try this, but I’m glad I did. I’m starting to feel like each of my hobbies is getting a fair amount of time per week. This has led to better moods and less anxiety that I’m “not getting anything done!” Do you have any productivity hacks for parents out there? Do I even have any regular readers? Oi.

Flint Project (Deep Learning + Architecture + Urban Planning)

This project is starting to pick up steam, so I need to make sure I’m ready with my part. If you recall, I’m hoping to put together a deep learning model for semantic segmentation of satellite (or aerial) imagery. Using this model, the aim is to identify unused or underutilized areas of land in the city of Flint, Michigan. One type of area that will be of interest is parking lots. My local expert (aka my brother) mentioned that when buildings are torn down in Flint, the lots are often paved over with asphalt and turned into parking lots. What would be a dream in downtown San Diego, where I live, is turning into missed opportunities in Flint. More to come, as we build this project over the Summer.

In order to create this model, I decided to focus on the relevant sections of the deep learning online class that I’ve been working on, rather than work through the material in a linear fashion. Despite that going against the way my brain works, with the above-mentioned lack of time, I think it will get me up and running more quickly. I will definitely go back and complete the material I missed at a later time. Or maybe not – maybe I find that this is an area that really interests me and I continue to focus my learning in this area. It seems like the benefits are myriad, though there isn’t as much buzz surrounding this field as, say, computer vision for autonomous vehicles. One thing that is certain, is that the fastai deep learning library is extremely accurate, can generalize well, and allows you to get up and running with minimal amounts of code. My next task is to understand the modules that I’ll need, in order to know what kind of data I need to input to the model. From there, I can train, validate and apply that model to this project. It seems like a tall order, but I’m confident that I can do it. Stay tuned.