Användning av maskininlärning för att välja ut porträtt
(2018)Computer Science and Engineering (BSc)
- Abstract
- This bachelor thesis evaluates the use of Clustering and Regression to separate good images from a given set of images. Whether the images are good or not is decided by the customers of the company that this system is developed for. In this thesis these customers are represented by a number of test persons. The images have been analysed with Amazon’s service Amazon Rekognition where the answers are used as parameters in the JAVA-based machine learning tool WEKA.
The main goal of this thesis was to find parameters that are capable of selecting good images and also to gain knowledge in machine learning and algorithms. The thesis contains results from several tests where both Clustering and Regression are tested.
The result of this bachelor... (More) - This bachelor thesis evaluates the use of Clustering and Regression to separate good images from a given set of images. Whether the images are good or not is decided by the customers of the company that this system is developed for. In this thesis these customers are represented by a number of test persons. The images have been analysed with Amazon’s service Amazon Rekognition where the answers are used as parameters in the JAVA-based machine learning tool WEKA.
The main goal of this thesis was to find parameters that are capable of selecting good images and also to gain knowledge in machine learning and algorithms. The thesis contains results from several tests where both Clustering and Regression are tested.
The result of this bachelor thesis is an evaluation that determine which of the methods Clustering and Regression, that are best suited to choose good images. The evaluation emerges into future areas of use for systems, like the system presented in this thesis. The result shows us that the learning capability of Regression is higher than it is for Clustering. Clustering on the other hand is more dependent of the ratio between number of parameters and quantity of data. The result also shows us that Clustering as well as Regression are able to choose good images with better results, than if they were randomly chosen.
the System
The system was developed in JAVA that only reached a development environment that managed to run the tests, meaning that no user interfaces was developed. The tests were executed from Eclipse.
Keywords
Clustering, Logistic Regression, Image Analysis, WEKA, Amazon Rekognition. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/8953239
- author
- Axelsson Ahl, David and Paulsson, Lotta
- organization
- year
- 2018
- type
- M2 - Bachelor Degree
- subject
- keywords
- clustering, logistisk regression, bildanalys, weka, amazon rekognition
- language
- Swedish
- id
- 8953239
- date added to LUP
- 2018-06-27 03:44:49
- date last changed
- 2018-10-18 10:38:37
@misc{8953239, abstract = {{This bachelor thesis evaluates the use of Clustering and Regression to separate good images from a given set of images. Whether the images are good or not is decided by the customers of the company that this system is developed for. In this thesis these customers are represented by a number of test persons. The images have been analysed with Amazon’s service Amazon Rekognition where the answers are used as parameters in the JAVA-based machine learning tool WEKA. The main goal of this thesis was to find parameters that are capable of selecting good images and also to gain knowledge in machine learning and algorithms. The thesis contains results from several tests where both Clustering and Regression are tested. The result of this bachelor thesis is an evaluation that determine which of the methods Clustering and Regression, that are best suited to choose good images. The evaluation emerges into future areas of use for systems, like the system presented in this thesis. The result shows us that the learning capability of Regression is higher than it is for Clustering. Clustering on the other hand is more dependent of the ratio between number of parameters and quantity of data. The result also shows us that Clustering as well as Regression are able to choose good images with better results, than if they were randomly chosen. the System The system was developed in JAVA that only reached a development environment that managed to run the tests, meaning that no user interfaces was developed. The tests were executed from Eclipse. Keywords Clustering, Logistic Regression, Image Analysis, WEKA, Amazon Rekognition.}}, author = {{Axelsson Ahl, David and Paulsson, Lotta}}, language = {{swe}}, note = {{Student Paper}}, title = {{Användning av maskininlärning för att välja ut porträtt}}, year = {{2018}}, }