Skip to content

Augmentation of a monocular image using a CNN generated depth map of the scene.

License

Notifications You must be signed in to change notification settings

acvictor/Monocular-Augmentation-Depth-Map

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Monocular Augmentation

This project deals with the augmentation of monocular images while attempting to maintain realistic occlusion and minimize manual intervention. The implementation uses Godard et. al. Unsupervised Monocular Depth Estimation With Left Right Consistency to estimate a dense depth map for an input image. Each texel is then rendered at the depth given by the depth map.

NMPC NMPC NMPC

Compilation and Execution Instructions

To compile and execute the code run

make run

Data

  • D.txt is a 768x1024 array of depth values
  • 1.jpg is the original image
  • result.png is the segmented image

Usage

  • Use the arrow keys to move camera up, down, left and right
  • Use the '+' and '-' keys to move camera in and out
    • '+' zooms in
    • '-' zooms out
  • Use left mouse button to rotate camera
  • Use 'b' to toggle bounding boxes
  • Use ESC to quit.

Attribution

Used https://github.com/mrharicot/monodepth to construct the depth map

About

Augmentation of a monocular image using a CNN generated depth map of the scene.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published