Yuxin Hou · Arno Solin · Juho Kannala
Codes for the paper:
- Yuxin Hou, Arno Solin, and Juho Kannala (2019). Unstructured multi-view depth estimation using mask-based multiplane representation. Scandinavian Conference on Image Analysis (SCIA). [preprint on arXiv]
MaskMVS is a method for depth estimation for unstructured multi-view image-pose pairs. In the plane-sweep procedure, the depth planes are sampled by histogram matching that ensures covering the depth range of interest. Unlike other plane-sweep methods, we do not rely on a cost metric to explicitly build the cost volume, but instead infer a multiplane mask representation which regularizes the learning. Compared to many previous approaches, we show that our method is lightweight and generalizes well without requiring excessive training. See the paper for further details.
Tested with:
- Python3
- Numpy
- Pytorch 0.3.0
- CUDA 9 (You can also run without CUDA, but then you need to remove all
.cuda()
in codes) - opencv
- imageio (with freeimage plugin)
To install imageio, run conda install -c conda-forge imageio
or pip install imageio
. To install the freeimage plugin, run the following Python script once:
import imageio
imageio.plugins.freeimage.download()
We provide our pretrained models of our MaskNet and DispNet to run the example code. Please download the models via the link
- Put both the model
masknet_model_best.pth.tar
and the modeldispnet_model_best.pth.tar
under the project folder. - Then just run the jupyter notebook file example.ipynb
This software is distributed under the GNU General Public License (version 3 or later); please refer to the file LICENSE
, included with the software, for details.