|4 ECTS credits|
Computer Vision is a very active research field with many interesting applications. Many of its recent successes are due to advances in Machine Learning research. It is therefore useful to study the two fields together and to draw cross-links between them.
The conferences with the strongest impact in Computer Vision are CVPR, ICCV, and ECCV, whereas NIPS and ICML have the strongest impact on the Machine Learning community. In this seminar we will discuss recent results presented at those conferences with a focus on the most interesting and innovative ideas. Participating students have the chance to get familiar with state-of-the-art solutions to problems in Computer Vision and Machine Learning and will get an insight into the involved techniques.
Successful participants will be awarded 4 ECTS credits.
- Seminar places are allocated in a central procedure. Registration for the seminar is only possible online via the registration page provided by the institute.
- The introductory meeting will be at the beginning of the semester.
- The seminar presentations will be held as a block seminar on 2-3 consecutive days shortly after the end of the summer lecture period (exact dates to be announced).
Master students: Bachelor degree Attendance of the lectures Computer Vision, Machine Learning, or Pattern Recognition and Neural Networks, or evidence of equivalent knowledge.
- Ethical Guidelines for the Authoring of Academic Work in German and English
- Declaration of Compliance in German and English
- LaTeX for the report (mandatory!): Template
- Slide template (optional): PPT, Keynote, LaTeX
- Structured Forests for Fast Edge Detection
- Decision Jungles: Compact and Rich Models for Classification
- Decomposing Bag of Words Histograms
- EVSAC: Accelerating Hypothesis Generation by Modeling Matching Scores with Extreme Value Theory
- Histograms of Sparse Codes for Object Detection
- Online Motion Agreement Tracking
- An Analytical Formulation of Global Occlusion Reasoning for Multi-Target Tracking
- Online Video Superpixels for Temporal Window Objectness
- Fast, Accurate Detection of 100,000 Object Classes on a Single Machine
- Visualizing and Understanding Convolutional Networks
- Piecewise Rigid Scene Flow
- Voxel Cloud Connectivity Segmentation) Supervoxels for Point Clouds
- Holistic Scene Understanding for 3D Object Detection with RGBD cameras
- Human Pose Estimation using a Joint Pixel-wise and Part-wise Formulation
- Multi-Source Multi-Scale Counting in Extremely Dense Crowd Images
- Crossing the Line: Crowd Counting by Integer Programming with Local Features
- Introductory Meeting: 24.03.2014
- Outline due: 12.05.2014
- Report due: 16.06.2014
- Slides due: 14.07.2014
- Presentations: 23.07.-24.07.2014