08-13, 12:00–12:40 (Asia/Yerevan), 114W PAB
The talk is about applications of active learning methods, mainly Monte-Carlo Dropout on 3D mesh/pointcloud semantic segmentation task. The topic is particularly interesting for practical applications of Deep Learning models on this type of data, as it gives a working approach for reducing the amount of data needed for training.
I will briefly go over the 3D mesh/pointcloud semantic segmentation task, and active learning, so that it's clear for the audience not familiar with this concepts. Then I will present the PointNet++ model architecture and Monte-Carlo Dropout approach, that are specifically used in the experiments. And finally I will share the experiment results with the audience. The audience is expected to have a hands-on knowledge in with ML and DL, to understand and be able to practically apply the ideas presented in the talk.
Previous knowledge expected
Erik Harutyunyan is a Machine Learning Engineer, mostly specialized and interested in the Computer Vision domain. He is currently working at SuperAnnotate AI as a Machine Learning Researcher and concurrently pursuing a Master's degree in Mathematics in Data Science at the Technical University of Munich.