You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: README.md
+29-4Lines changed: 29 additions & 4 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -58,15 +58,18 @@ So nothing fancy.
58
58
Are you tired of always searching for peoples website or the names of affiliations. Is if **TU Munich**, **Technical University Munich**, or **Technical University of Munich**? In the files `_data/affiliations.yaml`, `_data/authors_external.yaml`, and `_data/authors_internal.yaml` you store the information about you and your co-authors and their affiliations so you avoid mistakes in the future.
59
59
60
60
For authors, the scheme is
61
-
```
61
+
62
+
```yaml
62
63
{shortname}:
63
64
first_name: ...
64
65
last_name: ...
65
66
website: (optional)
66
67
is_me: true # to mark yourself
67
68
```
69
+
68
70
For affiliations it is:
69
-
```
71
+
72
+
```yaml
70
73
{shortname}:
71
74
long: ...
72
75
short: ...
@@ -82,7 +85,7 @@ This is where you should spend most of your time. Every publication is a subdire
82
85
83
86
`index.md`contains all the meta information about publications that is used in `_layouts/publications.html` to generate the project page. `_layouts/publications.html` itself uses smaller snippets from `_include` to generate a header, citation, and acknowledgements. We show an example file to introduce the different options
84
87
85
-
```
88
+
```yaml
86
89
layout: publications # defines the layout used for this publication. Leave it as publications or create your own layout under _layouts
87
90
permalink: /publications/pnec/ # defines the link used to publish the website
88
91
date: 2022_05_26 # determines sorting just take the date of the first publication as YYYY_MM_DD
acknowledgements: 'We thank our colleagues, especially Florian Hofherr, for proofreading and helpful discussions. This work was supported by the ERC Advanced Grant SIMULACRON and by the Munich Center for Machine Learning.' # (Optional) Will add a acknowledgements section on the project page.
135
138
```
136
139
137
-
The rest of the project page you have to define in the `content.html` file, where the template cannot help you anymore execpt to provide you with examples. More, maybe later.
140
+
The rest of the project page you have to define in the `content.html` file, where the template cannot help you anymore execpt to provide you with examples. More, maybe later.
141
+
142
+
### Includes
143
+
144
+
Includes are supposed to provide reusable templates to make content sections more consistent and easy to use. One example would be `comparison-slider.html`. This allows you to overlay two images with a slider reactive to the mouse. In can be used in the `content.html` file with
Jekyll allows to pass arguments to the includes that can be used in the `hmtl` code as you can see in the example. In this case, since a `.js` script is needed for this component, we have to add the script at the end of our `content.html` as well. Also be mindful, that the `slider-component.css` file is alread included in the `publications.html` file. (There should probably be a cleaner solution for this.)
161
+
162
+
If you want to add your own reusable components you can just add them to the `includes` directory. (Cleaning up would be important at some point.)
<i><b>Visual Odometry Results.</b> Rotation and translation error using KLT keypoints.</i>
138
138
<br><br>
139
-
The tables show the average results on the test set over 5 runs for SuperPoint and KLT tracks on KITTI, respectively. We show additional results in the supplementary material. Our methods consistently perform the best over all sequences, with the self-supervised being on par with our supervised training. Compared to its non-probabilistic counterpart NEC-LS, our method improves RPE<sub>*1*</sub> by 7% and 13% and the RPE<sub>*n*</sub> by 37% and 23% for different keypoint detectors on unseen data. Furthermore, it also improves upon methods that use weighting, like weighted NEC-LS and the non-learned covariances for the PNEC, significantly.
139
+
The tables show the average results on the test set over 5 runs for SuperPoint and KLT tracks on KITTI, respectively. We show additional results in the supplementary material. Our methods consistently perform the best over all sequences, with the self-supervised being on par with our supervised training. Compared to its non-probabilistic counterpart NEC-LS, our method improves $\text{RPE}_1$ by 7% and 13% and the $\text{RPE}_n$ by 37% and 23% for different keypoint detectors on unseen data. Furthermore, it also improves upon methods that use weighting, like weighted NEC-LS and the non-learned covariances for the PNEC, significantly.
0 commit comments