Nachricht

On Tuesday, May 7, Kai Timo Krüger will present the intermediate results of his master's thesis with the title:

Deep Learning-based SIFT-like Keypoint Detection for Object Contours in Real Images

Abstract:

The objective of this work is to develop a Convolutional Neural Network (CNN)-based keypoint detection approach for object contours in real images. A conventional detection method has been developed at GET Lab, but requires given object contours. Similar to Scale-Invariant Feature Transform (SIFT), a scale-space analysis is used to obtain scale and rotation-invariant keypoints and their characteristic scales.

In another work at GET Lab, a contour extraction method based on image segmentation to extract the object contours from real images has been developed. However, the sequential application of all substeps significantly increases the complexity of the overall process so that it is not real-time capable. A CNN-based approach integrating all substeps is investigated to address this and aim for a robust generalization.

Examples of existing CNN-based keypoint detection methods are Key.Net and SobelNet, which are not only real-time capable, but also achieve comparable results to traditional keypoint detection methods. However, they are not contour-based. In the context of this thesis, the complete process of contour-based keypoint extraction should be replaced by a CNN-based approach. The final approach should be executable as a single unit and real-time capable. The CNN should be trained in a supervised manner using images from the Segment Anything 1 Billion (SA-1B) dataset. The ground truth data, consisting of keypoint positions and their characteristic scales, will be generated using the detection method from GET Lab.