Hierarchical Annotation Tool
Status des Praktikums: offen
Betreuer:
Walter G. Kropatsch
Problemstellung
Problem formulation:A segmentation algorithm partitions an image into homogeneous regions
with respect to some low level cues (e.g. color).
A hierarchical segmentation method produces not only a single segmentation
but many, with each being coarser than the previous one.
An annotation tool should enable a human user to read the segmentation
produced by a hierarchical segmentation method and label the regions that
together form meaningfull objects (human, car, house)
by also allowing a hierarchical labeling of the parts
that make up the objects (head, torso, arms, legs, etc.).
Zielsetzung
Task of the Praktikum:
* Refine existing hierarchical annotation tool (in Java) that allows users to label regions in segmentation hierarchies.
* Extend annotation tool to automatically annotate new segmentation hierarchies based on labeled information gathered before (e.g. this region represents a human, automatically decompose and label the parts)
Gliederung
Schedule of the Praktikum:* Literature study (Segmentation, Segmentation Hierarchies)
* Implementation.
* Tests and Benchmarks.
