Effects of Video Compression formats on Neural Network Performance
Year: 2019
Type: dissertation
Abstract: The field of autonomous vehicles and driverless cars\nis a field which makes extensive use of machine learning and\nartificial intelligence, relying on it to make decisions. These\ndecisions require a vast amount of data in order to be properly\ninferred. This data is often in the form of images from a video\nfeed and an increase in the amount of data is directly correlated to\nmore refined decision making. What if you could remove certain\nparts of the data by compressing the image input, and how\nwould that influence the performance of the different machine\nlearning algorithms? In this report we have two datasets that\nwe compress in different ways. We then analyze the results\nof running a pre-trained neural network model on them, and\ncompare it’s performance to that of running the same neural\nnetwork model on the non-compressed datasets. The results\nshow that the removal of data via compression is not in a\nlinear relationship with neural network performance, and that\ndepending on the compression type, results may be favorable or\nunfavorable.
Authors: Marko Stanoevich, Jonathan Partain
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2995364994
