A simple package to reconstruct a 3D model of an object from a video taken for that object
This package uses Python 3.x
Also, you'll need OpenCV and numpy which can be downloaded using the following commands
pip install numpy
pip install opencv-python
To install modeler simply run this command from the command line:
pip install modeler
Or
pip install git+git://github.com/Wajdy759/modeler.git#egg=modeler
Then you can import it using:
import modeler
modeler has a module called SfM which handles wraps everything you'll need
sfm = SfM(results_dir, video_already_converted, video_path, video_sampling_rate)
Where:
results_dir
: the directory where the module outputs the .ply filevideo_already_converted
: a boolean set to true if the video used was already converted to images in 'input_images'video_path
: in case video_already_converted is False, the module uses the video in 'video_path' as inputvideo_sampling_rate
: the frequency at which the module extracts images from the video to use in the actual reconstructionTo actually run the algorithm and get the 3D model you'd use
sfm.find_structure_from_motion()
from modeler import SfM
sfm = SfM('results/', False, 'videos/vid1.mp4', 27)
sfm.structure_from_motion()
When the progrm terminates you'll get a set of .ply files in the folder named results
This project was heavily inspired by this article https://www.researchgate.net/publication/265190880_3D_reconstruction_from_multiple_images