The generation of trajectories with background

The generation of background

plot_video_segment.py is used to extract frames from video according to predefined frame rate.

Plot the trajectories with background

The key to plot trajectories on the background (frames) is to convert the location in the world coordinate to the pixel location via homography matrix.

import os
import numpy as np
import cv2


def world2image(traj_w, H_inv, background):
    # Converts points from Euclidean to homogeneous space, by (x, y) → (x, y, 1)
    traj_homog = np.hstack((traj_w, np.ones((traj_w.shape[0], 1)))).T
    # to camera frame
    traj_cam = np.matmul(H_inv, traj_homog)
    # to pixel coords
    a = traj_cam[2]
    traj_uvz = np.transpose(traj_cam/a)
    traj_uv = traj_uvz[:, :2]
    return traj_uv[:, :2].astype(int)

# Dataset name
dataset_name = 'zara2'

# load the homography matrix
H = (np.loadtxt(os.path.join(f'./{dataset_name}/H.txt')))
H_inv = np.linalg.inv(H)

# load background figure
background_image = cv2.imread(f'./{dataset_name}/889.png')
background_image = cv2.cvtColor(background_image, cv2.COLOR_BGR2RGB)

# convert to the pixel location
# observed_trajectory: Tx2 numpy array: represent the location of observed trajectories in the world coordinate
observed_pixels = world2image(observed_trajectory, H_inv, background_image) 

The script for plotting trajectories with background (frame) extracted from the video is plot_on_video.py.

Acknowledgement

The content largely borrows from OpenTraj benchmark.

@inproceedings{amirian2020opentraj,
      title={OpenTraj: Assessing Prediction Complexity in Human Trajectories Datasets}, 
      author={Javad Amirian and Bingqing Zhang and Francisco Valente Castro and Juan Jose Baldelomar and Jean-Bernard Hayet and Julien Pettre},
      booktitle={Asian Conference on Computer Vision (ACCV)},
      number={CONF},      
      year={2020},
      organization={Springer}
}

If you find this work useful in your research, please give a star to encourage me.