Shiaoming

check this function: https://github.com/Shiaoming/Python-VO/blob/c03e88fed01985a008eca14c8b49f8644fbf6d74/utils/tools.py#L38-L91 _Originally posted by @Shiaoming in https://github.com/Shiaoming/Python-VO/issues/7#issuecomment-755929333_

I mean to modify the path of the dataset in the configuration file. For more details about the dataset loader, you may refer to the dataset loader.

你显卡显存太小了：》

In my experience, superpoint+superglue is very robust.

之前也尝试过，但是没有训出来，后来就没再训练sift+superglue了。

1. [traingulate](https://en.wikipedia.org/wiki/Triangulation_(computer_vision)) the keypoints, the [opencv function](https://docs.opencv.org/3.4/d0/dbd/group__triangulation.html) might be helpful 2. and associate the same landmarks in the sequence

We will publish the training code for [ALIKED](https://github.com/Shiaoming/ALIKED) (the subsequent work of ALIKE, which we are working on). Then you can refer to the training code of ALIKED, they have...

Please refer to https://github.com/Shiaoming/ALIKE/raw/main/assets/ALIKE_code.zip.

it's probably the `rgbd_gt.txt` is incomplete. make sure the `rgbd_gt.txt` contains the correct file paths and gt poses like this: ``` # rgbd with groundtruth # timestamp rgb_filename depth_filename tx...

@zhanghua7099 I use the code below along with `associate.py` provided by TUM dataset to generate `rgbd_gt.txt`: ```python import argparse import os import sys LOCAL_PATH = './' if LOCAL_PATH not in...