Skip to content

Latest commit

 

History

History
145 lines (76 loc) · 9.38 KB

README.md

File metadata and controls

145 lines (76 loc) · 9.38 KB

CSSL Insights

🌪 A Rundown on CSSL Insights

CSSL Insights is a web application platform developed by Western Engineering research interns to support the Northern Tornadoes Project's capacity in collecting, analyzing, and sharing important research data in a meaningful way.

01 Sign-in Process

Users can create an CSSL Insights account using one of two methods: by entering their email address or by linking their Discord account. If using an email address, a verification email will be sent with a link to proceed. If using a Discord account, a pop-up will prompt the user to authorize the account.

02 Account Customization

Upon initial account creation, the user will be prompted to set their profile settings, including their name and profile picture (optional). Profile settings can be accessed at any time through the main dropdown in the top right corner of the screen.

03 Authentication

For full access to the CSSL Insights feature set, a user must be CSSL Authenticated. Any currently authenticated user can authenticate other users by navigating to Settings > Authenticate Users.

04 Main Dashboard and Navigation

The main dashboard provides quick access to the three CSSL Insights software tools. The navigation bar at the top persists on every CSSL Insights page and includes a main dropdown with access to all CSSL Insights pages and a button to configure the website theme.

CSSL Social

CSSL Social is a social media scrapping tool that utilized LLMs to scrape social media posts from Twitter, with the ability to add more sites. It will take these posts and grade them based on a classification model running on co:here that will determine the relevance of the post to the CSSL. The user can then view the posts and their grades, and can then choose to contact that user if future information is needed.

01 Social Dashboard

Social Dashboard

The dashboard is where you can view all of the current and past searches. You can also create a new search from here. The table also displays if the current search is enabled. If it is enabled, it will run based on the frequency defined in the search (every 6 to 48 hours).

02 Creating a New Search

Creating a new search is easy with the built in preset values. You can also choose to create a custom search with your own values. The search will be saved and can be viewed in the dashboard.

  • Search Name: The name of the search. This will be displayed in the dashboard.
  • Date Range: The date range of the search. This will be the date range of the posts that are scraped and how long the scraper will run.
  • Keywords: The keywords that will be used to search for posts. The scraper will search for posts that contain these keywords.
  • Negative Keywords: The negative keywords that will be used to search for posts. The scraper will not search for posts that contain these keywords.
  • Frequency: The frequency of the search. This will determine how often the scraper will run.
  • Max Results: The maximum number of results that will be scraped. The scraper will stop once this number is reached.
  • Location: Using the map marker, location based keywords will be added to the search. The scraper will search for posts that contain these keywords.

New Search

03 Viewing a Search

The search page is where you can view the posts that were scraped from the search. You can also view the search details and where posts are being found. The table of posts is where you will be able to view the posts and their scores. You can also view them on their respective sites.

View Search

04 Downloading Posts

You can download the posts from the search as a CSV or JSON file. This will download all of the posts that were scraped from the search given the time frame you select. Using the flag feature is helpful so that you can create code to either exclude or invert the categorization of the posts.

By downloading the posts, you can continue to train the ML model for the app or create validation sets for the model as well.

Download Posts

CSSL 360

CSSL 360 enables the uploading, viewing, and sharing of 360-degree panorama captures taken on severe storm damage surveys on-screen and in virtual reality (VR) using React Three XR (based on WebXR). Users can move through and around the street-level imagery and toggle between Before and After to compare changes that occurred from the storm.

01 360 Dashboard

The 360 dashboard allows users to create a new event path upload and access the table of pre-existing uploads.

02 Uploading New Event Paths

After providing event details such as the event name, folder name, and date, the upload process is broken into 3 distinct uploads: the framepos.txt file, the event panoramas, and the comparison panoramas.

Uploading the framepos file. CSSL 360 is designed around the NCTech iSTAR Pulsar camera workflow. Once data formation through NCTech Immersive Studio is complete, a framepos text file is created, storing all of the necessary geospatial data for each photo. CSSL 360 parses this data and uses it to plot map points, populate a Details pane alongside the 360 view, and more.

Uploading the event panoramas. After counting the number of data entries in the framepos.txt file, CSSL 360 determines the number of panoramas that it expects to be uploaded. Note that only numerically chronological file names, i.e., 0000000000, 0000000001, 0000000002, etc. are accepted.

Uploading the comparison panoramas. CSSL 360 references the latitude and longitude values for each panorama and uses the Google Street View API to fetch the closest available panorama for each of the uploaded iSTAR Pulsar panoramas. A copyable/downloadable list of panorama IDs is then given to be entered into Street View Download 360 for exporting. Note that each comparison panorama file must be named after its panorama ID.

03 Viewing Event Path Uploads (Screen and VR)

CSSL 360 uses Three.js to create a spherical texture map view of each panorama and CameraControls to look around the view. The following are the controls for the 360 view:

Alongside the 360 view is a Details pane that offers useful information about the panorama and more granular sequence controls. There is also an interactive map powered by the Mapbox API that plots a clickable point for every 5 images along the event route.

04 Sharing Event Path Uploads

From the 360 dashboard, users can copy a shareable public link that leads directly to its associated event path upload. An inline frame of the 360 view can also easily be embedded onto other webpages.

05 Additional Resources

Click here to learn more about using the iSTAR Pulsar camera and Street View Download 360.

CSSL LiDAR

Another way to capture visual data from a storm survey is in the form of LiDAR pointclouds. In conjunction with images from ground/aerial photography, LiDAR data can be used to identify failure modes of natural and man-made structures.

CSSL LiDAR is a way to quickly view LiDAR pointcloud scans without having to download a standalone program. Just upload your .las or .laz file and you're off to the races!

01 LiDAR Dashboard

LiDAR Dashboard

The LiDAR dashboard allows users to create a new scan upload and access the table of pre-existing uploads.

02 Uploading New Scans

LiDAR Upload Prompt

The upload dialog requires the user to upload the pointcloud file(s) in the *.las format or its compressed variant, *.laz. As a part of the upload process, metadata relating to scan/event dates, scan location, and scan type (e.g. ground vs. aerial) are required to be entered in.

After the upload process is completed, the backend automatically converts this to a binary format that the renderer can use.

Speaking of which...

03 Viewing Scan Uploads

LiDAR View

The LiDAR pointclouds are rendered using Potree Core which is a simplifed library of the venerable Potree software. As such, the core foundation of the renderer is built around THREE.js. To move the camera around the pointcloud, a combination of FlyControls and PointerLockControls is used to provide cursor and keyboard control.

Such controls are further detailed in the graphic below:

LiDAR Controls

04 Sharing Scan Uploads

Just like with the 360 dashboard, users can copy a shareable public link that leads directly to its associated LiDAR scan entry. An inline frame of the pointcloud view can also easily be embedded onto other webpages.