Product

Product Updates: June 2022 Roundup

Superb AI
|
4 min read
Want to explore more?
Sign up for an account to get started. No credit card required.

Introduction

Implementing scalable workflows remains challenging when building successful computer vision models. Still, the right features and tools can make the job much easier. Because of this, ML/CV engineers and labeling teams are always looking for ways to cut down on time, save money, and streamline their processes to prevent the common labor backlog that plagues so many companies.

At Superb AI, we always strive to provide the tools you need to accelerate your workflows, bolster your model’s performance and capabilities, and keep your data safe. With our most recent product release, we’ve implemented 2d cuboid annotation, added support for Azure integration, enhanced bulk operations within the SDK, and added auto-pause to your labeling tasks. Multi-factor authentication is also coming very soon (June 20th)!

Annotate 3D Objects in 2D Data with Cuboids

Bolster the efficacy of your autonomous vehicle datasets, heighten your robotics projects' accuracy, and improve the performance of your object detection models by using 2D cuboid annotation in place of standard bounding boxes. We’ve added this capability as a better way for computer vision models to interpret objects captured with a 2D camera lens.

Cuboids are drawn by clicking the draw tool, forming a single-faced square, and then extending each vertices to connect at the corners of another square, essentially creating a box. This tool allows you to draw outside of the canvas to demonstrate parts of an image that were not captured by the camera. Cuboid annotation helps a model perceive the depth of an object and its distance with enhanced precision compared to other labeling types such as the aforementioned bounding boxes.

Click here to learn more.

Upload Your Data Faster and Easier with Azure Blob Storage

The Superb AI labeling platform now supports Microsoft’s object storage service, Azure Blog Storage. This allows you to use your Azure service as a data source without manually downloading and uploading them (once integrated). And will enable you to further protect your data by optionally allowing read-only access to your containers and data assets, meaning your data remains within your storage, and we don’t create any copies of your data. Coupled with other new features like multi-factor authentication, you can rest easy knowing your data is always safe when using any aspect of the Superb AI Suite.

Click here to learn more.

Complete Labeling Projects Faster with New SDK Commands

Organization and effective collaboration can make or break your labeling project and is often the difference between completing it on time or days/weeks later. We’ve added new bulk operations to our SDK that help streamline communication and collaboration between your ML engineers and labeling teams. With new commands, your ML engineers can assign and unassign labelers and reviewers, change a label’s status, and edit relevant tags. This added functionality was built to help you save time and interact with your labeling team more easily without navigating to the UI version of our labeling platform. Each command is generated by adding tags through a line of Python code and is very simple to execute.

Click here to learn more.

Keep Better Track of Your Team’s Labeling Progress with Autopause

Labeling a dataset isn’t always a linear process, as each image or frame can take a varying amount of time. For example, a simple image with a few bounding boxes may take seconds, while a complex segmentation task can take minutes. In addition, labelers are people, and people get side-tracked. Failing to make room for distractions when tracking label completion times can paint an inaccurate picture.

We understand that these metrics help data project managers and others make staffing, workflow, and budgeting decisions. To facilitate more accurate tracking, we’ve implemented a new autopause functionality. If a user is inactive for more than 30 seconds, the timer at the top of each label will pause. In addition, 30 seconds will be subtracted from the overall time to account for the period in which a labeler was inactive before the timer stopped.

Trust Your Data is Secure with Multi-Factor Authentication (Coming Soon)

On June 20th, multi-factor authentication (MFA) will also be added to the labeling platform to provide an extra layer of trust. In addition to being SOC 2 compliant, MFA generates a unique six-digit code at each login so that your data is protected by more than just a password. Backed by Amazon Cognito, Superb AI’s MFA process helps ML practitioners and labeling teams protect their data with just a single and easy extra step.

In a later release, we'll also be adding the ability for users to have their devices remembered for a more seamless login experience. For remembered devices, a project manager can choose to have their team’s devices always set to be remembered, never retained, or have an individual user opt-in. The same goes for tracking. This functionality grants each team and user flexibility in how they add this layer of security to their projects.

Click here to learn more.

What This Means for Your Team

New additions to the Superb AI Suite translate into better productivity and efficiency levels for your labeling operations. Enhanced communication through the SDK, heightened security with MFA, the introduction of cuboid annotation, and the ability to autopause can directly and positively impact your workflows and improve trust in your data privacy. Feel confident in your model’s performance with 2d cuboids, and make more informed decisions on operations and staffing with more accurate metrics. 

With each update added to the Superb AI Suite, our goal is to improve your data labeling experience, ultimately helping you get to production AI faster. Click here to hop back in or sign up for free to get started today.