A python implementation of Lucid Data Dreaming.
Lucid Data Dreaming is a data augmentation technique for semi-supervised video object segmentation, which is proposed in Lucid Data Dreaming for Multiple Object Tracking, A. Khoreva, R. Benenson, E. Ilg, T. Brox and B. Schiele, arXiv preprint arXiv:1703.09554, 2017.
To generate a pair of images, you can refer to
We firstly generate the background image and then used it to
do the Lucid Data Dreaming. The former invokes
patchPaint.py and can be done in about one minute. Once the background is generated,
there is no need to do the same work again. The Lucid Data Dreaming
lucidDream.py, using only around 0.4 seconds to generate a pair of images, which is much faster than the matlab version.
(The time mentioned above is on a server with NVIDIA TITAN X GPU and Intel(R) Xeon(R) CPU E5-2680 v4 @ 2.40GHz)