mkocabas

set smooth

## Nanorc files
include "/usr/share/nano/nanorc.nanorc"

## C/C++
include "/usr/share/nano/c.nanorc"

## HTML
include "/usr/share/nano/html.nanorc"

mkocabas

How to visualize the body part segmentation?

This script lets you to visualize the body part segmentation labels of SMPL, SMPL-H, and SMPL-X body models. Follow the instructions to be able to run this script.

Instructions

1. Download the body models you would like to visualize.
2. Install the requirements
3. Run python visualize_body_part_segmentation.py <body_model> <body_model_path>

mkocabas

####### LOSS FUNCTION #######
class MultivariateGaussianNegativeLogLikelihood(nn.Module):
    def __init__(self):
        super(MultivariateGaussianNegativeLogLikelihood, self).__init__()

    def forward(self, pred_mean, pred_var, gt):
        mu = pred_mean
        logsigma = pred_var

        mse = -0.5 * torch.sum(torch.square((gt - mu) / torch.exp(logsigma)), dim=1)

mkocabas

import torch
import numpy as np
import neural_renderer as nr
import matplotlib.pyplot as plt

from smplx.body_models import SMPL

def main():
    device = 'cuda'

mkocabas

import cv2
import numpy as np
import math

cx = 88 #principal point x coord
cy = 109 #principal point y coord
w = 178 #image width
h = 218 #image height
near = 10 #near plane
far = 20 #far plane

mkocabas

import os
import json
import shutil
import subprocess
import numpy as np
import os.path as osp

def run_openpose(
        video_file,
        output_folder,

mkocabas

import dgl
import torch
import torch.nn as nn
import time


def build_pose_graph():
    g = dgl.DGLGraph()

    g.add_nodes(16)

mkocabas

import numpy as np
import torch

def compute_similarity_transform(S1, S2):
    '''
    Computes a similarity transform (sR, t) that takes
    a set of 3D points S1 (3 x N) closest to a set of 3D points S2,
    where R is an 3x3 rotation matrix, t 3x1 translation, s scale.
    i.e. solves the orthogonal Procrutes problem.
    '''

mkocabas

import subprocess
import os
import os.path as osp

def extract_video(vid_filename, output_folder):
    cmd = [
        'ffmpeg',
        '-i', vid_filename,
        f'{output_folder}/%06d.jpg',
        '-threads', '16'

mkocabas

import torchvision
import os.path as osp
import numpy as np
import cv2
import torch
import time

def get_video_opencv(fname):

    vid_tensor = []