2020-10-23
New features and changes as of 2020-10-23

Issues

Issue Membership on the Frame Detail View

When looking at the detail view for an image or sensor frame, you'll now see a list of all issues that the image or its labels are associated with.
The image detail view will now show issue membership info.
Issues that the entire image / frame is associated with.
If individual labels are associated with issues, they'll be shown alongside a count indicator.

Color Embeddings by Issue

The embedding view can now color by issue membership!
Points colored by the type of issue (if any) they are associated with.

Issue Element Counts

On both the issue list and issue detail page, you will now be able to see the total number of elements associated with each issue.
Element count on the issue list.
Element count for a specific issue.

Renaming Issues

You can now rename issues after creation.
Issue names now have an edit button
Enter in a new name and click save!

Issue Management in the Python Client Library

The python client library now has support for managing issues. A more in-depth tutorial will be coming soon -- in the meantime, feel free to reach out for more details!
1
# List Issues
2
issue_manager = al_client.get_issue_manager(project_id)
3
issues = issue_manager.list_issues()
4
​
5
# Get full content of an issue
6
full_first_issue = issue_manager.get_issue(issues[0]['id'])
7
​
8
# Delete an issue
9
issue_manager.delete_issue(issues[0]['id'])
Copied!

General UI Improvements

Toggle-All for Embedding Coloring

You can now quick-toggle to enable or disable all categories when looking at categorical data in the embedding view. You can also shift-click to enable or disable ranges.
Click "All" to quick toggle all of the categories

Infinite-Scroll Progress Indicator

Queries or confusion matrix drill-downs that produce infinite-scroll views of the data will now show the total returned results, as well as what portion of it is currently on screen.
At the bottom of infinite scroll views, you'll now see where you're at.

Sensor Modality Support + Asset Hosting

Upload Assets Directly in the Python Client Library

Sometimes, you just have some files on disk that you want to load into Aquarium, without thinking about hosting and uploading. The python client now supports directly uploading files, returning a URL that you can use as usual:
1
uploaded_url = al_client.upload_asset_from_filepath(
2
project_id,
3
dataset_id,
4
/path/to/file.png
5
)
6
frame.add_image(..., image_url=uploaded_url)
Copied!

Support for per-sensor and per-label coordinate frames in 3D data

Robotics applications often have multiple sensors and inputs, which might exist in their own coordinate frame. For content rendered in 3D (point clouds, 3D label cuboids, etc.), each element can be associated with a custom coordinate frame consisting of a position and orientation.

Direct support for PCD point cloud formats

In addition to KITTI style binary arrays for lidar data, we also directly support loading PCD data (ascii, binary, and binary-compressed) in the browser app.

Python Client support for Audio Data

Audio sensor input was previously supported, but is newly exposed via the python client library.
​
Last modified 1yr ago