mihaibujanca

# MP4 video, RGB only
gst-launch-1.0 filesrc location=output.mp4 ! decodebin name=dec ! queue ! videoconvert ! autovideosink dec. ! queue

# RGB-D video

# Realsense RGB-D + IMU streaming 

mihaibujanca

# Authors: Mathieu Blondel, Vlad Niculae
# License: BSD 3 clause

import numpy as np


def _gen_pairs(gen, max_iter, max_inner, random_state, verbose):
    rng = np.random.RandomState(random_state)

    # if tuple, interpret as randn

mihaibujanca

Running main() from gtest_main.ccUsing Opt v0.2.0
nUnknowns = 	54
nResiduals = 	0 + (@parametersSym).DataG.N * 3

nnz = 	0 + (@parametersSym).DataG.N * 48

compile time: 	1.1069540977478
problem plan complete
GPU memory usage: used = 514.000000, free = 3530.437500 MB, total = 4044.437500 MB
14269.380859375

mihaibujanca

Running main() from gtest_main.ccUsing Opt v0.2.0
s/Opt/API/src/o.t:1887: terra cost($idx : int32,$P : ProblemParameters) : float
s/Opt/API/src/o.t:1888:     var $midx : int32 = $idx
s/Opt/API/src/o.t:1851:     var $r1 : float_3
                            var $r2 : float
                            var $r3 : float_8
                            var $r4 : float
                            var $r5 : float
                            var $r6 : float
                            var $r7 : float

mihaibujanca

Project Report

Project repository | Commits list | PR to OpenCV

Brief

The aim of this project was to implement DynamicFusion, published in 2015 by Newcombe et al., which introduces an RGBD SLAM system capable of reconstructing 3D models of non-rigidly deforming scenes. The method represents the scene as a canonical model and a warp field that transforms the model at each frame, in order to distinguish between already observed data and data that needs to be added to the model. The system employs KinectFusion [Newcombe et al., 2011] as a 3D reconstruction engine, with an altered pipeline:

mihaibujanca

// Use Gists to store code you would like to remember later on
console.log(window); // log the "window" object to the console