Advanced Topics In Computer Vision And Deep Learning

Spring 2020

Requirements

Prerequisites:

The course is intended to cover topics of research from the past year. Basic knowledge is assumed.
Students who took the Intro to Vision course can cover on their own the topics below that were not taught.

Deep Learning basics and applications. (Stanford's CS231n course covers it all), In particular, familiarity with:
- Neural network optimization- Back propagation, SGD variants (Example: ADAM)
- Losses and Layers: (Examples: Softmax, Batch-norm, Cross-entropy-loss)
- CNNs: Basics, applications, architectures (example: ResNet), Batch-Norm. [video1], [video2]
- RNNs: Basics, LSTM, applications, combinations with CNNs (Example: basic image captioning). [video]
- GANs: Basics, applications, domain transfer (example: pix2pix) [video] (The last part of the video)
- Basics of Deep Reinforcement learning: Shallow familiarity with basic methods is sufficient (Example: policy gradients)
  [video]
Classical Computer VIsion basics. (Basic course at the faculty covers it all)

Course Requirements:

Attend all the lessons. In case of miluim, illness or other justified absence please let us know.
Read for each class the assigned reading material.
Prepare a presentation.

Grades will be given based on:

Quality and clarity of presentation.
Level of understanding of the material and the ability to present in a clear and simple way.
Active participation in the other lessons.
Fulfillment of reading assignment and attendance.