BoulderSpot: Switzerland

Model results updated Oct 8th!

BoulderSpot is a bouldering app that uses computer vision and satellite imagery to help you find new boulder problems.

Just click on the Discover button to view satellite imagery marked with model-found boulders, pull their coordinates, then go climb!


About the Project

Note: it may take a couple of seconds for the app to load.

BoulderSpot uses a convolutional neural network with millions of parameters to learn image/computer vision tasks from a training dataset. A dataset of about 15,000 images was used to train the model, and now it can classify whether or not an aerial photo of Switzerland could be a potential bouldering area or not. The dataset is based on current great outdoor bouldering areas and sampled images of terrain throughout the country.

Above is an interactive demo of the computer vision model that classifies whether a boulder is a climb_area or other. Please note:

Models in the Streamlit demo are more compact than the final model implemented due to file size limits.The final model was trained on a binary labelled dataset of ~15,000 images.

Discover New Boulders

The Add Data to Map screen may take a minute - Desktop works better.

Above: a visualisation of potential bouldering areas found on Swiss aerial imagery by a trained CNN model. Some data validation visits to the potential boulder areas are documented on the blog page.

tl;dr: look for clusters of large blue dots.

Potential boulder areas found by the model are partially prefiltered in the map above: potential boulders (tiles) need 2+ adjacent boulder areas and a class probability of 80% to be visible. Filters can be changed in the app menu as required.The bluer the marker is, the more confident the model is that the area is good for bouldering.The larger the marker is, the higher the ratio of adjacent image tiles that are also classified as potential bouldering areas.A search bar and all hiking trails in Switzerland (data source: SwissTopo) are available in the full-screen map as an optional layer. See usage tips for details.The visualisation and search is still in-progress! As of Oct 8, only ~13% of the country has been searched by area - see details and visualisation here.

Contact

Questions, comments, feedback, or want to help?


Thank You

Your note has been sent and will be responded to in due time.

GDPR Disclaimer

Disclaimer: Privacy & Others

Last updated: July 26, 2021

Interpretation and Definitions

Interpretation

The words of which the initial letter is capitalized have meanings defined under the following conditions. The following definitions shall have the same meaning regardless of whether they appear in singular or in plural.

Definitions

For the purposes of this Disclaimer:

  • Company (referred to as either "the Company", "We", "Us" or "Our" in this Disclaimer) refers to BoulderSpot.
  • Service refers to the Website.
  • You means the individual accessing the Service, or the company, or other legal entity on behalf of which such individual is accessing or using the Service, as applicable.
  • Website refers to BoulderSpot, accessible from boulderspot.io

Disclaimer

The information contained on the Service is for general information purposes only.

The Company assumes no responsibility for errors or omissions in the contents of the Service.

In no event shall the Company be liable for any special, direct, indirect, consequential, or incidental damages or any damages whatsoever, whether in an action of contract, negligence or other tort, arising out of or in connection with the use of the Service or the contents of the Service. The Company reserves the right to make additions, deletions, or modifications to the contents on the Service at any time without prior notice. This Disclaimer has been created with the help of the Disclaimer Generator.

The Company does not warrant that the Service is free of viruses or other harmful components.

External Links Disclaimer

The Service may contain links to external websites that are not provided or maintained by or in any way affiliated with the Company.

Please note that the Company does not guarantee the accuracy, relevance, timeliness, or completeness of any information on these external websites.

Errors and Omissions Disclaimer

The information given by the Service is for general guidance on matters of interest only. Even if the Company takes every precaution to insure that the content of the Service is both current and accurate, errors can occur. Plus, given the changing nature of laws, rules and regulations, there may be delays, omissions or inaccuracies in the information contained on the Service.

The Company is not responsible for any errors or omissions, or for the results obtained from the use of this information.

Fair Use Disclaimer

The Company may use copyrighted material which has not always been specifically authorized by the copyright owner. The Company is making such material available for criticism, comment, news reporting, teaching, scholarship, or research.

The Company believes this constitutes a "fair use" of any such copyrighted material as provided for in section 107 of the United States Copyright law.

If You wish to use copyrighted material from the Service for your own purposes that go beyond fair use, You must obtain permission from the copyright owner.

Views Expressed Disclaimer

The Service may contain views and opinions which are those of the authors and do not necessarily reflect the official policy or position of any other author, agency, organization, employer or company, including the Company.

Comments published by users are their sole responsibility and the users will take full responsibility, liability and blame for any libel or litigation that results from something written in or as a direct result of something written in a comment. The Company is not liable for any comment published by users and reserve the right to delete any comment for any reason whatsoever.

No Responsibility Disclaimer

The information on the Service is provided with the understanding that the Company is not herein engaged in rendering legal, accounting, tax, or other professional advice and services. As such, it should not be used as a substitute for consultation with professional accounting, tax, legal or other competent advisers.

In no event shall the Company or its suppliers be liable for any special, incidental, indirect, or consequential damages whatsoever arising out of or in connection with your access or use or inability to access or use the Service.

"Use at Your Own Risk" Disclaimer

All information in the Service is provided "as is", with no guarantee of completeness, accuracy, timeliness or of the results obtained from the use of this information, and without warranty of any kind, express or implied, including, but not limited to warranties of performance, merchantability and fitness for a particular purpose.

The Company will not be liable to You or anyone else for any decision made or action taken in reliance on the information given by the Service or for any consequential, special or similar damages, even if advised of the possibility of such damages.

Contact Us

If you have any questions about this Disclaimer, You can contact Us:

Model - Information

This page contains details about the model used in BoulderSpot, including the confusion matrix, stats, and more.

Model:
resnext101_32x8d trained for 20 epochs on a binary classification task, with a labelled dataset of ~15,000 images.The labelled dataset includes images of classic bouldering locations in Switzerland (Cresciano, Magic Wood, etc.) as examples of the positive class.Train/test split was 70/30, respectively, and images were augmented during training (blur, crop, rotate, etc.).


Key Stats

Confusion Matrix

Class Probability Distribution - Holdout Set

Learning Rate Optimization

Examples of Class Predictions - Holdout Set

Positive Class (climb_area)

Negative Class (other)

Usage Tips - Kepler.gl Viz

The Kepler visualization is an overlay on top of the open-source Kepler.gl library. Essentially, you can zoom in pretty far to validate the potential bouldering areas the model has found.


Open/Close Menu Panel

The viz takes a couple of seconds to load the data - while it does this, you will see the below screen. It will then update the map.

When it loads, you can close the menu panel with the small arrow at the top - it can be hard to notice, especially on mobile:


Hiding Blue Boulder Markers

After finding what might be a nice boulder, the overlay squares can be hidden by opening the menu panel with the same arrow, then by clicking on the eye icon by the respective layer:


After the location is confirmed to be nice, write down the coordinates or head over to the data export page to get more info. You can get this information by clicking on the squares to pull up the tooltip.


Swiss Hiking Trails

It is easy to visualize the approach to any given potential bouldering area by looking at the hiking trails to get there. All hiking trail data is made available through SwissTopo's swissTLM3D, a large-scale topographic landscape model of Switzerland.

Open the full-screen version of the map and see below.

Adding in the hiking trail data is easy. Open the same menu panel, and click on the eye icon on the layer Swiss TLM3D Hiking Trails - English to enable visualization.

The trails are broken into the following categories (also shown in the legend):

1. Hiking trail
2. Mountain hiking trail
3. Alpine hiking trail
4. Other

Note that the map display in-site does not contain trails due to file size restrictions. To see all the trails, click on full screen on the Discover page or click here. It is recommended to do this on a desktop due to the number of objects loading, etc.

BoulderSpot - Export Data

Use the below to find details on a bouldering area that looked appealing on the maps.

Details
The HTML file below is interactable for sorting, filtering, etc. If there are any issues, you can download the data as .xlsx here.An easy way to find a point spotted on the Kepler map is the NGD_ID field as these are unique per area.Distance stated is linear distance from Zurich HB as a proxy for how far it is. If you're reading this, you know that the actual travel time depends on which mountains you're going around.Once decided on a spot to check out, enter the lat/long data provided in the Lat/Long column into Google or your map provider of choice, and voilà!Climb outside.


Linked below the table is another interactive plot that goes into further detail.


EDA on BoulderSpot Data


Use the below to find details on the boulderspot data - distance from HB, Kanton, so on.


Search Status


Switzerland has a lot of land area (~41,285 km^2), correspondingly many aerial photographs (~2 TB of data). It will take some time for the model to work through the important areas, so if your favourite crag does not show up, there is a decent chance the model has not been evaluated on it yet.

The below is a visualization of what area has been searched and what area remains. While not the snappiest map ever, it is possible to zoom in and search.

The total area scanned is 5360.05 km^2
The total area of Schweiz is 40690.155 km^2
scanned 13.17 % of the country

*Note: the above is from the code itself, which may deviate slightly from actual areas due to some lakes being removed and other small differences due to projection system and so on.


Last update: Data current as of Oct 6, 2021

Blog / Trips

Data Validation Trip 03 - A return to Valhalla

The first time I went (see: Trip 01), we ended up on top of the cliff (see pictures) and then it rained. This time the goal was to get some climbing in - we succeeded.

This place is massive, and while I don't want to claim it's bigger than Magic Wood, the area is huge. Upon further research, it seems to have formed in the Goldau Landslide in 1806.

A repost of the map in Trip 01 to save scrolling:


Data Validation Trip 02 - "outpost"

We went to Glarus to see another boulder area found by the model. This one was quite high elevation-wise, but the approach was reasonable as a cable car takes you up there.


What did it look like on the BoulderSpot Discover page?

outpost-from-map

Approach Trail Map

The location of the cable car in Ennenda is also shown below.

Data Validation Trip 01 - "valhalla"

The first trip! I did not climb due to rain (and the cliff I took pictures from), but it looks quite promising. This one is in Schwyz, by Goldau (45 min train from HB).

You can see how massive the boulder field is from the photos in the slideshow.
Text


What did it look like on the BoulderSpot Discover page?

outpost-from-map

Approach Trail Map