Skip to content

shade-117/KNN-projekt

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Projekt KNN

Téma: Odhad absolútnej hĺbky z obrazu (na prírodných dátach)

Autoři: Adam Ferencz, Tomáš Mojžiš, Petr Mičulek

Useful links

Project structure

Some parts of this project are taken from other repositories:

Other files are scripts implemented by us. Main scripts are:

  • geopose/train.py - training script, containing also evaluating on test dataset
  • geopose/dataset.py - script containing DataLoader class
  • geopose/losses.py - script containing implementation of loss functions
  • and other minor utility scripts

Notes

Run from root repo directory (KNN-Projekt).

Loading JPEG uses TurboJPEG library that must be installed through a package manager.

sudo apt install libturbojpeg

Python package dependencies can be installed from requirements.txt.

pip install -r requirements.txt

How to launch

  • install requirements.txt with pip
  • for training on datasets/geoPose3K_final_publish run python geopose/train.py
    • set the parameters inside the script
  • for plotting run python plot_predictions.py - set the weights_path variable
    • it could not be passed as parameter because of megadepth model builder uses argparse and it interferences

About

Projekt KNN - Single image depth estimation

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages