Auflistung nach Autor:in "Haenselmann, Thomas"
1 - 2 von 2
Treffer pro Seite
Sortieroptionen
- TextdokumentAutomated Annotation of Sensor data for Activity Recognition using Deep Learning(INFORMATIK 2017, 2017) Benndorf, Maik; Ringsleben, Frederic; Haenselmann, Thomas; Yadav, BharatWithin this work-in-progress, we aim to automate the annotation of Sensor data for generating training data for Activity Recognition (AR) of multiple persons. Usually, the activities are executed and recorded from test persons under the supervision of an instructor, which may influence in many cases the natural behaviour of the test persons and the authenticity of the data. In this work, we suggest how this influence can be reduced and how the Sensor data can be annotated automatically by using video capturing, openpose for extracting human key points and a neuronal network to classify the activities. By automatically annotating the selected activities we show the feasibility of our approach.
- KonferenzbeitragForking Agents in Sensor Networks(Informatik 2005 – Informatik Live! Band 2, 2005) Haenselmann, Thomas; Effelsberg, WolfgangInformation about local events are to be distributed by agents in a sensor network. The agents are sent out traveling hop-by-hop into random directions, leaving a path of information behind them in the nodes visited. An arbitrary node requesting for information also emits an agent traveling into a random direction. Though only a fraction of all nodes will contain the requested information the high probability of a requesting agent intersecting the path of an informing colleague is exploited to find the information. A major problem of this approach called Rumor Routing is that it produces a highly non-uniform distribution of information. We have tackled this problem by introducing Forking Agents.