Date of Award
Campus Access Thesis
Master of Science (MS)
Over the course of the last nine years we have truly seen a machine learning revolution, which fueled computer vision to new horizons. New methods and new datasets are released at a rapid pace, in order to take advantage we need projects and pipelines that will be able to handle these advancements. Datasets are only getting bigger and bigger, computer scientists have a lot of responsibility for making pipelines efficient, fast and scalable using the latest available technology. It is a balancing act, usually we either choose efficiency but sacrifice scalability or the other way around, so it is very important to aim for the golden mean. Semantic segmentation is one of the key applications in computer vision and I have built a pipeline that addresses all aforesaid concerns at the same time, resolving real world semantic segmentation problems. It is highly modular, easy to use and works even with highly sizeconstrained datasets. My pipeline provides various preprocessing tools as well as training and testing code with a lot of additional functions to use and try. At its core it is highly parallelized and scalable from a single CPU core to multiple, and from one GPU to many.
Zhurkevich, Alexander, "EFS: An Efficient, Fast and Scalable Semantic Segmentation Pipeline Using Tensorflow 2" (2021). Graduate Masters Theses. 719.
Free and open access to this Campus Access Thesis is made available to the UMass Boston community by ScholarWorks at UMass Boston. Those not on campus and those without a UMass Boston campus username and password may gain access to this thesis through resources like Proquest Dissertations & Theses Global or through Interlibrary Loan. If you have a UMass Boston campus username and password and would like to download this work from off-campus, click on the "Off-Campus UMass Boston Users" link above.