• EN

Practical Introduction to Deep Learning for Computer Vision

Supervisors: Dr. Ralf Krestel and Christian Bartz

Organisation: We are planning to have pysical meetings if possible. If necessary, we will switch to a completly virtual setup using BBB.

We will have a Q&A session on Oct. 28th at 15:15 using BBB (https://bbb.hpi.uni-potsdam.de/b/ral-1ib-9my-ol2).
Interested students can register by sending an email to ralf.krestel@hpi.de until Oct. 29th.
If more than 9 students want to participate, we will choose randomly and notify the chosen ones on October 30th.
The introductory, first session will then take place on Wednesday, November 4th at 15:15 in F.E-06 (and, if needed, on BBB).

This seminar gives a practical introduction to three common computer vision problems which we will tackle using current state-of-the-art deep learning approaches. Students will work in teams to understand and then extend/adapt existing solutions. We collaborate with the Getty Research Institute in Los Angeles, which provides training and test data, namely, a large photo collection.

Learning Objectives

Students will learn to...

  • solve classical computer vision tasks
  • employ deep learning architectures and algorithms
  • read and understand scientific publications
  • read code and extend and adapt existing code
  • conduct experiments and evaluate different solutions
  • present their results
  • write a report about their experiments

Schedule

Time: 15:15
Location: Campus 3, F-E-06 and BBB https://bbb.hpi.uni-potsdam.de/b/ral-1ib-9my-ol2

Date Topic
October 28th Q&A session only BBB
Before October 30th Register by email (ralf.krestel@hpi.de)
October 30th Email notification whether you are accepted
November 4th Organization & Introduction to Image Classification
   
December 2nd Team Presentations & Introduction to Object Recognition
   
January 6th Team Presentations & Introduction to Image Retrieval
   
February 3rd Team Presentations & Closing of Seminar
March 15th (tentative) Hand-in Written Team Reports

(still updated, subject to change)

Grading

  • 20% Image Classification Project
  • 20% Object Recognition Project
  • 20% Image Retrieval Project
  • 20% Presentation
  • 20% Written Report

(subject to change)

Prerequisites

  • Basic knowledge of machine learning is necessary (What is cross-validation? What is an objective function? What is classification? What are labels? etc.).
  • Having prior knowledge about deep learning is an advantage but not required (What is backpropagation? What is a CNN? etc.).
  • Having some experience with Python is recommended, as we will use Pytorch for the seminar.

Contact

This course is organised by Dr. Ralf Krestel and Christian Bartz and we collaborate with the Getty Research Institute, which provides the data.

References