Wednesday, September 25, 2019

SITP Student Francis Akenkora works on Fusing OpenROV for Subsea Infrastructure Assessment.

I can still recall the day I became interested in this project. One of my friends came to me informing me that his professor was interested in adding additional undergraduate students into his lab with the purpose of fusing an OpenROV for subsea infrastructure assessment. I immediately became interested in when Dr. Zhang told us he wanted to see if it would be possible to leverage the OpenROV to use object detection methods based on machine learning. I’ve dealt with problems regarding detection and estimation throughout my academic career as an EE in both the classroom and through my capstone project, but neither allowed me to explore some of the hot button topics within the field. Using some of the fancy new ultra-fast detection methods such as YOLO (You Only Look Once) and SSD (Single-shot Detection) have convinced me that machine learning works! While machine learning does in fact work, it’s not to say that it is without hiccups. One thing I learned is that like humans, machines need a lot of exposure in order to learn. Trying to train a neural network can be hard especially if there aren’t any publicly available data sets (or any data sets for that matter) regarding what you’d like your network to be able to detect.

On a weekly basis, I would say most of my time is spent reading. I knew that there was a lot of information I needed to understand before I could truly appreciate the research I was conducting, but I didn’t anticipate there to be this much reading! For just about every problem I could’ve asked myself, there was thankfully a research paper that managed to answer that question, even if it wasn’t directly related to my research. Borrowing ideas from others has been an eye-opening experience to the wealth of knowledge that is readily available.

One thing this project has taught me is that I really do love soling problems and finding more problems to solve. I hope to pursue a career that manages to blend the interests I’ve acquired throughout my academic career. I love SONAR and I love image processing as well, so I hope to pursue a career where I can develop systems that use SONAR to generate images. If I do decide to pursue a graduate degree, this is the field I would become heavily interested in.