We seek a PhD student to research motion computation and object detection under occlusions, smoke, fog, low image contrast, and with tiny as well as fast-moving objects.
This research is important because such phenomena and objects occur frequently in natural environments. Think about valuable insects to be tracked and counted, or trees to be monitored as part of recreation programmes. Especially forests and open land are becoming more and more important in vision applications. In contrast to cities and the artificial environment, current approaches to motion computation and recognition are very much challenged by the natural environment and it is currently not clear if models mainly developed for the autonomous driving case can be applied in these scenes.
Furthermore, occlusion is a longstanding problem in vision and it is still not satisfyingly solvable under the paradigm of neural networks. How visual learning itself can help to overcome the occlusion problem is still debated and determined, but there are good arguments that more than visual learning is needed to solve the problem.
The research will use a new methodology by combining traditional, bottom-up methods from first principles with neural networks and ML. You will establish potential training and benchmark data and define appropriate performance metrics for different sensor modalities, such as visual and thermal.
You are part of a team working at the cutting edge of technologies for AI systems by addressing practical problems from various stakeholders of European research projects. You will gain experience from renowned international scientists and build your network to build up your scientific profile in cooperation with the European research community. Your work could be under the scientific examination of Prof. Daniel Cremer at the Technical University of Munich.
If you are interested in this position, please receive more details and apply here. The position is open until it gets filled.