1 / 16

Learning Generic and Person Specific Objective Functions

This Diplomarbeit provides an overview of model-based image interpretation, generic objective functions, and learning objective functions. It discusses the traditional approach, ideal objective functions, and experimental evaluations. It also explores person-specific objective functions and their challenges.

tresam
Télécharger la présentation

Learning Generic and Person Specific Objective Functions

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Learning Generic and Person Specific Objective Functions Diplomarbeit

  2. Overview • Model-based Image Interpretation • Generic Objective Functions • Traditional Approach • Ideal Objective Functions • Learning Objective Functions • Experimental Evaluation • Person-specific Objective Functions • Experimental Evaluation Technische Universität München Sylvia Pietzsch

  3. Model-based Image Interpretation • ModelContains a parameter vector p that represents the model‘s configurations. • Objective FunctionCalculates how well a parameterized model fits to an image. • Fitting AlgorithmSearches for the model that fits the image best by minimizing the objective function. Technische Universität München Sylvia Pietzsch

  4. Traditional Approach • Designer selects salient features from the image and composes them. • Based on designer‘s intuition and implicit knowledge of the domain. • shortcomings: • time-consuming • resulting objective function is not ideal Technische Universität München Sylvia Pietzsch

  5. Ideal Objective Functions P1: Correctness Property:The global minimum of the objective function corresponds to the best model fit. P2: Uni-Modality Property:The objective function has no local extrema or saddle points. Technische Universität München Sylvia Pietzsch

  6. Example: Comparing Objective Functions a) image b) along perpendicular c) edge values d) designed objective function e) ideal objective function f) training samples g) learned objective function Technische Universität München Sylvia Pietzsch

  7. Learning the Objective Function (1) Technische Universität München Sylvia Pietzsch

  8. Learning the Objective Function (2) Technische Universität München Sylvia Pietzsch

  9. Learning the Objective Function (3) 6 styles · 3 sizes · (5 · 5) locations = 450 features Technische Universität München Sylvia Pietzsch

  10. Evaluation 1: Used Features • Model trees tend to select the most relevant features. • Edge-based features are hardly used at all. Technische Universität München Sylvia Pietzsch

  11. Evaluation 2: Robustness Indicators measure the fulfillment of P1 and P2: I1: Correctness Indicator Distance between the ideal position of the contour point and the global minimum of the objective function I2: Uni-Modality Indicator Total number of local minima divided by the size of the considered region Technische Universität München Sylvia Pietzsch

  12. Evaluation 3: Learning Distance Technische Universität München Sylvia Pietzsch

  13. Person Specific Objective Functions • Single ImagesThe objective function has to take any appearance of a human face into consideration. ➱ moderate accuracy • Image SequenceThe appearance of a person‘s face only changes slightly. • Consider particular characteristics of the visible person, e.g. beard, glasses, bald head,... ➱ increase of accuracy • Challenges: • Learn specific objective functions for groups of persons offline. • Detect the correct group online. Technische Universität München Sylvia Pietzsch

  14. Evaluation: Fitting Results 45 persons from news broadcasts on TV Technische Universität München Sylvia Pietzsch

  15. Outlook • Learning objective functions for 3D-Models • Integration of further image features • Compute the image features on the fly • Automatic detection of the visible person:e.g. via AAM parameters Technische Universität München Sylvia Pietzsch

  16. The End Technische Universität München Sylvia Pietzsch

More Related