Diminished reality (DR) aims at removing objects automatically in live video streams, in a way which is unrecognizable to human observers. Approaches working on mobile devices utilize image inpainting algorithms to fill the holes (target region) of the removed objects in a plausible way.
We provide five video sets recorded with a Microsoft Surface 2 Pro handheld device. The video sets are located in directories named “setX” with X holding the video number. Each set consists of the input frames, the keyframes with the detected planes used for inpainting and the output of our published DR algorithm. These images are located in the directories “images”, “DR/dr_keyframes” and “DR/dr_out”. All stored frames have a resolution of 1280x720 pixels.
Moreover, each DR directory comprises two files: “KeyColor.txt” and “planeFinderConfig.ini”. The first file defines the target region generation using its color values and basic morphologic operations. The second file specifies the parameter of the plane extraction via iterative RANSAC (details are in the paper).
Before recording the videos, the handheld’s camera was calibrated. The results can be found in the “calibrationFile720p.xml”, which is located in the root directory. The root directory also includes the license file (CC BY-NC 4.0) and a more detailed introduction text file.
Lastly, we also recorded the accelerometer- and the gyroscope- data of all camera captures (files: accel.txt and gyro.txt)
The videos are provided as one compressed and password protected zip file. The password is mmk@dr.
If you use the diminished reality videos in your research, please cite our paper:
P. Tiefenbacher, M. Sirch, and G. Rigoll, Mono Camera Multi-View Diminished Reality", in Proc. WACV, IEEE, 2016