computer vision system for our autonomous vehicle. I mostly wrote the code on my own, and I got very creativity in using uint16 instead of float to shave off some valuable seconds of the processing time. However, I didn’t quite have the focus to comment my code appropriately, even though it was highly efficient code. The end result is that my code was not used by anyone else, which I regret.
At Notre Dame, I quickly learned my motivation was to graduate on time. This meant I had to focus. I determined when I wanted to finish, and I back tracked all the time requirements of each module of my work. The result was that I had a rough idea of when things had to get done, and this allowed me to stay more or less on track.
At DSC, I had a mixed bag. When I started, we were only using the entire face (as opposed to regions) for face recognition when the literature at the time suggested any method would be improved by multiple regions fused together. My manager had other things for me to work on, but I kept getting drawn to this low hanging fruit. To even do the testing, I revamped the code and sped it up by a factor of 10. I was using the oldest computer at the company, so my code needed to run just a bit faster if I wanted to do the experiments properly. I threw together a few rough regions, and I was able to quickly show how much more improvement one got form multiple face regions. After that, my previous tasks went out the window, and while I was happy, I’m sure it was frustrating to my manager.
1. Fundamentals of AI, ML and Deep Learning for Product Managers
2. The Unfortunate Power of Deep Learning
3. Graph Neural Network for 3D Object Detection in a Point Cloud
4. Know the biggest Notable difference between AI vs. Machine Learning
On the other hand, when I came up with a method for improve the 3D depth quality by upsampling, I spent two months of my time implementing the method in C++ that was unnecessary. I had been trying to implement the more complex methods of upsampling (based on the griddata function in Matlab) when in fact, I had found major gains using the least complex. I was stubborn in part because of my curiosity on how to efficiently implement these higher order methods. Finally, I turned my work full force on getting the basic method implemented, and I was able to achieve a certain level of focus.
At Apple, on the Watch, my creativity lead to a breakthrough in background heart rate. Two months later, I received feedback about the previous year (before this breakthrough) that I was not focused enough. I couldn’t deny the claim. I had been working on Wrist Detection, and certainly I had times where I went off on a tangent. However, my argument had been the saving grace of my creativity. I would still argue that my creativity was worth a lack of focus, and I would go so far as to say I could not do both.
On Face ID, that changed in part because of the volume of tasks and task switching. I was compelled to find a better way to track progress on different tasks and also not drop any of them. This started with a Notepad document that turned into a Word document, and that morphed into a Numbers sheet. I was able to add in my ideas of curiosity into my sheet so I would not forget them, but I could also see the other work I had to do and prioritize accordingly. Suddenly, I was able to stay focused and switch focuses quicker just by knowing where I was on a project. This didn’t seem to dampen my ability to be highly creative in trying to solve problems either.
I’m still working on the balance, and I believe this is key to hitting a maximum potential. I have ideas all the time, but that doesn’t mean they are all good. It just means I have so much to throw against the wall, one of them is bound to stick.
