Electrical Engineering

Welcome to the Master Thesis Presentation by Jacob Justad
Titel: Self-Supervised Fixed-Scene Adaptation for Object-Detection in Real-Time Surveillance: A Comparative Study of YOLO11 and RF-DETR
Datum: 20 januari Tid: 10.00 Rum: E2 room 3364 EDIT-rummet Opponent: Tommy Liu Examinator: Jennifer Alvén
Abstract: This thesis, conducted in collaboration with Axis Communications, investigates self-supervised fixed-scene adaptation for real-time object-detectors in an edge-computing surveillance context. While modern object-detectors achieve strong results on general-purpose benchmarks, deployment in static camera scenes introduces distinct challenges: domain shift to a specific viewpoint, limited availability of scene-specific labels, and stringent compute and memory budgets on-device. At the same time, the stationary background of surveillance footage provides exploitable structures, as do their temporal dependencies of video-frames.
This study conducts a comparative analysis of two state-of-the-art object detection architectures: the Transformer-dominant RF-DETR and the convolutional neural network (CNN)-dominant YOLO11. The thesis employs the 100Scenes dataset to represent a broad range of surveillance environments. By systematically varying model scales, adaptation strategies and environmental conditions the experimental design yields more than 3400 distinct runs.&nbsp; The study further investigate background-context fusion (BF) and an assessment of a proposed &nbsp;on-device self-supervised labeling strategy based on SAHI combined with bi-directional ByteTrack. Furthermore, the work examines the extent to which smaller, specialized models can match or approach the performance of substantially larger models and how these configurations react to seasonal domain shift.
Experimental results demonstrate that RF-DETR consistently achieves higher accuracy, smoother convergence, and greater robustness than YOLO11, albeit with higher hardware demands. In contrast, YOLO11 variants (with a frozen backbone) leverage the larger trainable capacity of the neck and head to enable high scene-specific adaptability. While this yields significant gains under quality labeling, it increases sensitivity to imperfect pseudo-labels and the risk of overfitting. The proposed self-supervised labeling strategy provided reliable performance gains across all architectures by recovering hard negatives—specifically small, occluded, and low-confidence instances. Conversely, while Background-Context Fusion consistently improved RF-DETR, it proved inconsistent for YOLO11 and failed to increase robustness against seasonality, suggesting it induces background-dependent overfitting. The study also shows that small, specialized models can compete with larger general models. Further, the adapted models exhibit a decrease in relative performance compared with the non-adapted models during a seasonal domain shift.&nbsp;

Welcome!
Jacob and Jennifer

Welcome to the Master Thesis Presentation by Tommy Liu
Titel: Exploring Image-to-Text Visual Search Using Open Source Models
Datum: 20 januari Tid: 11.00 Rum: E2 room 3364 EDIT-rummet Opponent: Jacob Justad Examinator: Ida Häggström
Abstract: Visual searching refers to the use of visual data, typically images, in order to perform a search rather than textual input. Most visual search implementations rely on performing similarity searching over image features, in which a user-submitted query image is compared against all searchable entries’ features before returning sufficiently similar results. This thesis explores a different method which utilizes image descriptions generated by vision-language models instead of image features, where the descriptions are converted into embeddings in order to match with other search entries. Evaluation data indicate that the method can provide satisfactory retrieval performance in addition to maintaining a low search query execution time, provided that an adequate vision-language model is employed and sufficient server capacity is available.

Welcome!
Tommy and Ida

Architecture and Civil Engineering

Bachelor’s thesis Chemistry and Chemical Engineering

Chemistry and Chemical Engineering

Communication and Learning in Science

Computer Science and Engineering

Industrial and Materials Science

Life Sciences

Mathematical Sciences

Mechanics and Maritime Sciences

Microtechnology and Nanoscience

Physics

Space, Earth and Environment

Technology Management and Economics

Presentation of master thesis

Choose subject

In order to play this video you must open your cookie settings and allow external media cookies.

In order to use basic functions like page navigation and access secure areas of the website, you must open your cookie settings and allow necessary cookies.

In order to use this service you must open your cookie settings and allow external services cookies.

To help us understand how visitors interact with the website you must open your cookie settings and allow statistics cookies.

Course syllabus for

Subject

Each academic year contains different courses

See programme page at universityadmissions.se

Educational area

Permissible substitutions of compulsory courses

To see your programme syllabus – select the academic year in which you started the programme.

[en: page.studyprogrammepage.syllabusesview.coursesdescription]

Presentation of master thesis

Master thesis presentations at Electrical Engineering

20 January, 2026

10.00-11.00 Jacob Justad, MPDSC

11.00-12.00 Tommy Liu, MPDSC