In many modern vision and graphics applications, deep neural networks (DNNs) enable state-of-the-art performance for tasks like image classification, object detection and segmentation, quality enhancement and even new content generation. In this talk, we will demystify basic concepts behind DNN training: from the definition of a deep neural network to critical parameters controlling deep learning model training. We will examine problems typically encountered when training a model and will share best practices for their mitigation. Finally, we will shed some light on commonly used software frameworks.
Getting Started with Vision AI Model Training
Session Speakers
-
Ekaterina Sirazitdinova
Data Scientist, NVIDIA
Ekaterina is a data scientist at NVIDIA who specializes in solving computer vision and video analytics problems by means of AI. Her focus also includes fast deep learning inference on embedded devices. Previously, Ekaterina was a research engineer at Konica Minolta Laboratory Europe, where she developed deep learning algorithms for medical image analysis. She has authored several peer-reviewed journal and conference publications on various applications of image-based reconstruction, localization and tracking. Ekaterina received her Ph.D. in Computer Science and M.Sc in Media Informatics from RWTH Aachen University in Germany, she also holds a Diploma in Business Informatics from BSUEL in Russia.
