Name: TensorFlow Lite, Complex Convolutions and Deep Face Recognition
Start: 2020-12-20T19:00:00+02:00
End: 2020-12-20T20:30:00+02:00

Agenda:
19:00 - 19:05: Opening words
19:05 - 19:25: TensorFlow Lite - Elisheva Guahnich
19:25 - 19:45: Complex Convolutions - Idan Tobis
19:45 - 20:05: ArcFace: Additive Angular Margin Loss for Deep Face Recognition – Daniela Gingold

-----------------------------------------------------------------------------------------------

TensorFlow Lite - Elisheva Guahnich
-----------------------------------------------------------------------------------------------
For mobile devices, microcontrollers and embedded systems, Tensorflow Lite is a good solution for running the Deep Learning Models with a nice performance and portability.

Elisheva Guahnich is a machine learning specialist, project manager and software developer with more than 20 years of experience. She is an experienced consultant who has worked in the health, insurance, and financial sectors for both large and small commercial companies and government agencies. Elisheva is an expert in Data Analysis, Business Intelligence and Machine Learning.

Complex Convolutions - Idan Tobis
-----------------------------------------------------------------------------------------------
Many real-world signal sources are complex-valued, having real and imaginary components. Surprisingly, convolutional neural networks can be relatively easily tweaked to handle complex valued and complex weighted data. This lecture delves into the details of such tweaking.

Idan’s expertise is making complicated things look simple. Having an extensive and fruitful career in establishing and managing R&D teams for startup companies in Israel and abroad, and after leading a successful exit deal in North America , Idan has vast experience in realizing new and innovative ideas, from a sketch on a napkin to a working product in the market, along with an extensive patent portfolio of 34 patents to his name.

ArcFace: Additive Angular Margin Loss for Deep Face Recognition – Daniela Gingold
-----------------------------------------------------------------------------------------------
The most widely used multi-сlass сlassification activation function is SoftMax. However, does it explicitly maximize class separability? In this talk, we will present an ArcFace approach, which obtains highly discriminative features for face recognition, its effectiveness, and its easy implementation.

Daniela is a Computer Vision and Deep Learning Algorithm Engineer, currently researching a drug design for a novel antibiotic. She holds an M.Sc. in Technology Management and is on her way to another M.Sc., in Computer Science.

Lea Polachek

Primrose Machine Learning Meetups

Technology

Artificial Intelligence

Machine Learning

Artificial Intelligence Applications

Machine Intelligence

Artificial Intelligence Programming

Machine Learning Interpretability

Deep Learning