MMER_Locate

Example of MMER_Locate Result

Since our research focusses on the final target application of automatic Object Analysis, a reliable visual localization and tracking of the corresponding object is mandatory.


In case of Face Analysis (FEASy), the quality of the analysis is highly dependent on a precise initialization of our algorithm. Thus, a fast and robust head localization is not sufficient but only the exact positions of the eyes provide an accurate starting point for subsequent steps. However, the final analysis is highly computationally expensive, which limits the time and resources for the localization and demands high efficiency from the MMER_Locate technology.


We therefore enhanced the common cascaded filter based localization algorithm, which requires immense amounts of training images and time for statistical modeling. In exchange, the computational effort in the application phase is negligible at remarkable localization reliability.

The MMER_Locate technology is fully integrated in MMER_Lab and constitutes the starting point for all high level analyses as developed for MMER_FEASy.