Theses

If you are a student at UHH interested in a thesis with our group, you can contact Dr. Christian Wilms. Below is a list with selected titles of B.Sc. and M.Sc. theses completed in our group to orient you towards potential topics. You can also see our recent papers for possible directions for a thesis.

Selection of titles of complete Master theses:
- A Deep Learning Approach for Top-down Attention with Attribute Preference
- Salient object detection with AttentionMask
- 3D Segmentation in the Context of Inscriptions
- Active Visual Object Search Using Reinforcement Learning
- Saliency-Guided Sign Language Recognition
- Object Discovery in 3D Scenes via Shape Analysis using Adapted PCLV
- Learning Efficient Deep Feature Representations for Indoor Visual Positioning

Selection of titles of completed Bachelor theses:
- Weakly Supervised Object Detection in RoboCup Scenarios
- Lokalisierung von Flugzeug-Leitwerken
- IoU Predictions for Segmentation Mask Propsals
- Klassifikation von malignen Melanomen mittels konditionierten ConvNets
- Segmentierung von Schienenbildern zur automatisierten Wartung
- Bildklassifikation von Flugzeugtypen
- Segmentation of numerical weather prediction data for characterization of atmospheric airmasses
- Object Detection in Remote Sensing Image Data Using AttentionMask
- TileAttention: Entdeckung sehr kleiner Objekte
- Superpixel Pooling for Instance Segmentation

Necessary requirements:
You should have some pre-knowledge in computer vision before starting a thesis. For a B.Sc. thesis you should have at least attended the lecture "Einführung in die Bildverarbeitung", the lab course "Praktikum Computer Vision", or have equivalent knowledge. For a M.Sc. thesis, you should have attended the lectures "Computer Vision 1" and "Computer Vision 2", and ideally also the master project, or have equivalent knowledge. Pre-knowledge in machine learning is also helpful.