My scientific interest lies in visual motion analysis and tracking and in the importance of motion perception in object recognition and scene understanding. I currently focus in my research on the problem of fragmented and dynamic occlusion in machine and human vision.
Fragmented and Dynamic Occlusion
Please move your mouse pointer (or finger) over the image and click the left mouse button (or touch your display) to start the animation and to perceive motion.
Motion reveals a little car through a fence ...
... or a fragmentally occluded person through foliage.
Motion makes a camouflaged person in thermal images visible.
Object Detection
Motion helps a detector to aquire online a robust representation (green circles) of a car starting from an initial bounding box (WACV'14, CVPR'15).
Vehicles are reliably detected in satellite images by using motion (arXiv'22, CVWW'20).
Simultaneous Localisation and Tracking (SLAT), Camera Calibration
Motion of a person gives clues for the calibration of multiple security cameras with non-overlapping fields of view (
DAGM'10,
IEEE TPAMI,
CVPR'10,
CVPR'11,
ICCV'11) ...
... as well as structure from motion captured by an additional moving camera (
AT Patent,
3DV'13,
ECCVW'14).
Further Examples
Motion let us perceive depth and many more things! For more examples, visit D. Heeger's webpage.
My research profile is under construction.