dataset

dataset

This folder contains the skeleton of the crop dataset used to train LoFTR and do the crop alignment task.

It also includes scripts to download addition data, and prepare the datasets for training LoFTR.

*Note: The datasets here do not contain original images or generated depth images. Please download them seperately. The code will use the images and depth images by their absolute paths, and sometimes create symbolic links to them.

Folder structure

• crops
  • train: Dates from 2018 used to train LoFTR
  • Wheat_2019: Experiment V.B in the paper
  • exp_pipeline: Experiment V.C in the paper
• crop_indices: Indices necessary to train LoFTR. Generated by generate_indices.py. Here we only present an example. Remember to generate a new one with your own path of the downloaded dataset.
• download_dataset.py: Script to download images in the dataset. Refer to here for its usage.
• generate_indices.py: Script to generate crop_indices. Refer to here for its usage.