HyperLabel, a new desktop data labeling application for Machine Learning (ML) just announced by Sixgill, offers the fastest path to creating high-quality labeled datasets for better ML models.
With HyperLabel, there’s no need to upload files to an external service. Users retain complete ownership, privacy and control of their data, while accelerating project onboarding and completion with quick and easy usability anchored on the desktop. It’s all cloud-free, highly scalable and locally installed.
HyperLabel is designed to be fast, easy and accurate, from setup to label export. Customizations are straightforward and explanations are clear. HyperLabel will even use ML itself to give labeling projects a speed and accuracy boost with pre-trained models for common objects that will automatically create labels for you.
An easy QA interface will enable quick and efficient batch reviews of labeled data, to further streamline and simplify the labeling process. This will allow developers, engineers and data scientists to spend less time labeling and more time training their ML models.
Because HyperLabel is so easy to use with the power of ML, labeling projects will be more error free. HyperLabel enables accurate labeling that is critical to the success of ML models and the applications they inform. Labeling errors or inconsistencies damage the quality of training datasets, inhibit model performance and are causing many ML projects to fail.
With disruptive new tools that save time, cut costs and increase user control with a cloud-free, locally installed implementation, HyperLabel upends the assumption that accurate labeling is inevitably tedious and slow.
HyperLabel is available now for download via the Mac App Store or Microsoft Store. Valuable new features will be included in early August. Additional versions of HyperLabel will launch in the coming months, accommodating the full spectrum of data labeling needs, from individual developers to large teams:
- Developer, available now, is free to get started and includes the first 3,000 labels created.
- Developer Unlimited, beginning on August 12, will give users unlimited labels for only $9.99 per month after they surpass the first 3,000 labels. Optionally, users can select a pre-paid annual subscription for only $99.
- Pro, when released later this year, will offer powerful enhancements and advanced capabilities such as ML-assisted object tracking and pre-trained architectures, cloud collaboration, import/export support and labeling of 3D data types including DICOM, LiDAR.
- Enterprise, available now, will meet the needs of organizations with large and distributed data science teams; licenses will be custom-quoted based on scope.
“For our own customer projects, we need fast, high-quality data labeling to build ML solutions, including vision models for object identification in video data,” says Logan Spears, Innovation Chief at Sixgill.
“None of the available labeling tools gave us what we needed, so we built HyperLabel, and it worked so well that we had to share it. Developers shouldn’t have to trade speed for quality and ease of use.”
HyperLabel will deliver these powerful features and benefits:
Data privacy & control: With HyperLabel, you’ll always retain complete ownership, privacy and control of your data. Keep it private and label it where it lives. You are not forced onto the cloud.
Simple user experience: Get from project setup to label export in a few easy steps. Label objects without unnecessary clicks. See how much is left to do, and if the file you’re viewing has already been labeled.
Scalable: HyperLabel can handle even the most complex data labeling projects and can be used by solo developers, up to large teams. Or, request HyperLabel Managed Labeling Services and let HyperLabel experts do it for you.
Easy access: HyperLabel puts everything you need right at your fingertips. No need to upload files to an external service. Simply import your files from a hard drive or connect to your cloud storage and label away!
Easy export: Export labels to JSON, COCO, Pascal VOC, YOLO and other common formats, and include them in your training process.
Flexible user-defined schemas: Configure label schemas by selecting from rectangles, polygons, point, feature points, free text, select and multi-select, for almost any use case.
ML-automated labeling: HyperLabel uses the power of ML itself with pre-trained classifiers for predictive labeling of common objects. Advanced ML integration is coming soon.
Confident QA: The Quality Assurance (QA) interface will let you perform quick and efficient batch reviews of labeled data to ensure quality and accuracy.
Saves money: Labeling projects can sometimes take weeks or months and cost tens of thousands of dollars. HyperLabel optimizes and accelerates the process, so labeling gets done faster.