Two months ago, we announced the launch of GT Studio, a powerful web-based data labeling platform that has one of the best free plans for ML teams. The platform and all its premium features are completely free for a team of 5 users.
In this article, we’ll show you how GT Studio can help you set up projects and workflows quickly, manage annotation teams, and label data seamlessly on a single platform. Let’s dive right in!
For best results: Sign up here and follow this guide to start labeling immediately.
Take a quick tour of the platform with Akshay, Product Head at Playment
GT Studio offers both mature labeling tools and project management software to help you optimise your labeling operations. In this video, watch Akshay create a simple labeling project in less than 5 minutes.
The best part about this platform is that you don’t need to be an expert engineer to start using it, anyone can start labeling data using GT Studio. It’s as easy as it can get.
GT Studio: Main Components and Functions
The Acronym GT in GT Studio stands for Ground Truth. The platform is divided into two main interfaces GT Manage and GT Create. Understanding how these interfaces operate can largely simplify the bigger chunks of labeling activities ML teams are often tasked with.
1. GT Manage: Better Data Pipeline Management Software
To put it simply, GT Manage is essentially a management interface within GT Studio that lets you create your workspace, set up labeling projects and workflows, invite teammates, control how they interact with data, upload data batches, and assign labeling tasks. You can also upload pre-labeled data and access robust analytics on the same interface. Project management ends up hogging a lot of time with other platforms and that’s one thing we’ve focused on eliminating with GT Studio.
2. GT Create: High-Precision Labeling Tools
GT Create is the interface that actually allows you to create annotations or add labels to your datasets. The annotation interface is heuristically designed to support complex use cases. With intuitive tools and simple UI, the platform’s core features aim to reduce the amount of time annotators spend on labeling without sacrificing the quality of the annotations.
Setting Up A Labeling Project On GT Studio
You can configure a project and start labeling your data in just 4 easy steps.
Step 1: Create A New Project In Your Workspace
Once you log into GT Studio, you can create a new project in your workspace by simply entering details like Project Name and Base Type Annotation for the project. You’ll be the default owner, but you can add more teammates and assign it to anyone that’s a part of your workspace.
Base type annotation is the primary annotation that is required for the project. The selections in GT Studio currently include different types of 2D image and video annotations.
Step 2: Configure Classes & Attributes
The next step is to define the annotation specifications that are required for your project. These include classes and attributes. Class is essentially the object of interest in a given image/frame and attributes refer to the additional information pertaining to the object class.
For example, a Car is an object class and it could have multiple attributes like states [stationery, moving], occlusion [low, medium, high], etc.
Step 3: Choose A Pre-Built Workflow [or set up a custom workflow]
The next step is to set up a workflow for your project — this determines how your data flows from one stage of the process to the next. Based on the most commonly used workflows, we’ve offered four main workflows that you can choose from:
- Maker Workflow: The tasks are directed to the maker and then prepared for QC or export.
- Maker-Editor Workflow: The tasks undergo an additional check at the editor level where all editing and feedback tools are automatically enabled.
- ML-Proposals Enabled Workflow: The tasks have ML proposals pre-enabled where annotators can correct auto-labeled annotations instead of creating new ones.
- Classification Workflow: The tasks are prepared for classification and QC or export.
You can edit these workflows to create your own custom workflows, even for pre-labeled data.
Step 4: Drag and Drop Data [or use our powerful APIs]
The next step is to upload your data onto the platform for labeling. You can simply drag and drop the zip files into the dashboard or use our powerful APIs that allow you to transfer data using your own models. We support all commonly used data formats. Once the data is uploaded, each image/frame is created into a labeling job automatically.
Creating Annotations Using GT Studio
The simple interface allows you to easily create annotations based on the settings of your project. The tool is intuitive and easy to use. There are many keyboard shortcuts that allow you to label data faster and built-in QC features that ensure accuracy.
Managing A Labeling Project On GT Studio
The custom project dashboard provides all the important details about the progress of your project.
Robust productivity, quality, and mistakes analytics help you analyze annotator performance and streamline your labeling pipeline by ensuring your annotators are being productive.
Granular productivity details like man-hours spent on labeling, number of annotations created or edited, making time, editing time, etc provide a better picture of the overall team and individual performances.
Quality metrics like precision, recall, accuracy, annotations created, etc help you maintain better quality practices for labeling data.
Mistake classification allows you to check parts of the process where your team/annotators are losing time because of mislabeling data. Granular details like mislabeled classes, attributes, and incorrect labels will help you improve your labeling instructions to save more time for future labeling projects.
There’s a whole lot more!
From importing pre-labeled data for quality control to using your own models to ingest data into GT Studio, there’s really a lot you can do with the tools and controls we offer in GT Studio. We’ve designed it to make the annotation process convenient and fast so you can do your best work with zero hassle.
The platform is powerful and completely free for a team of 5 users.
Want more detailed information about setting up your first project? Check out our detailed documentation at ~ https://playment.gitbook.io/gt-studio/