Internship Report in the Multimedia Research Lab at University of Louisville (6 months)
Image annotation , Image Segmentation, Feature Extraction, Multiple Instance Learning, Bag of Features.
Image Annotation based on partially labeled data consists in assigning labels to new-to-the system images based on learning data in which the labels are bag-level only. The image annotation in this work is both image-level and region-level. The groundtruth data is only image-level labeled. The image annotation requires image segmentation, feature extraction and Multiple Instance Learning Classification.
The process for image Annotation Algorithm is summarized in this flowchart!
Some results of the Image annotation algorithm: