Show simple item record

dc.rights.licenseCC-BY-NC-ND
dc.contributor.advisorPoppe, Ronald
dc.contributor.authorVeroni, Eleni
dc.date.accessioned2024-02-15T14:50:00Z
dc.date.available2024-02-15T14:50:00Z
dc.date.issued2024
dc.identifier.urihttps://studenttheses.uu.nl/handle/20.500.12932/45943
dc.description.sponsorshipUtrecht University
dc.language.isoEN
dc.subjectFace detection in complex environments, particularly in the context of monitoring preterm infants, presents unique challenges due to occlusions such as hats or medical equipment, variation in lighting, video quality etc. Standard face detection models often fall short in such scenarios. Addressing the occlusion problem, this thesis proposes a novel two-part YOLOv5 model approach. The first part includes training the model on a customized dataset with bounding boxes of different facial parts. The
dc.titleUncovering the Invisible: Improving Face Detection Under Occlusions
dc.type.contentMaster Thesis
dc.rights.accessrightsOpen Access
dc.subject.keywordsface detection, preterm infants, AI, CNN, YOLO, YOLOv5, explainability, gradcam, hirescam
dc.subject.courseuuArtificial Intelligence
dc.thesis.id23035


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record