Over the last few months, Hasty has seen continued, accelerating growth. To keep the momentum going and deliver on our promise of built for and by engineers, Hasty is pleased to announce Hasty Pro Plans.
This article is about what you can expect from our pro plans, how you can test all features for free when you start your pilot project and how you only pay after you have proven Hasty’s value when you scale for production.
Our mission is to get your project into production faster than ever before. How? By helping you to create hundreds of thousands of annotations in the blink of an eye—resulting in a large, clean, completely custom training dataset.
We built our next-gen annotation tool around the following three principles:
Increasing the speed and quality of your annotation process. For this, we built a set of AI-assisted tools that actively learn to automate the majority of your work.
Allowing you to integrate Hasty with your workflow. We designed a tool for the way you work and not a tool that requires you to change the way you work.
Giving you control over your budget and users. Our costs are transparent and granular role-management features allow you to organize your team the way you need to.
Hasty comes armed with assitants to help you label image data faster than any other tool on the market. What makes us different in comparison with other tools is that they use a generalized model to pre-label data, our models are trained bespoke on the data you provide. This allows us to create accurate, customized assistants that adapt to your individual use-case. The purpose of our assistants is to automate annotation work to the maximum extent.
Hasty actively learns as you go, meaning the automation and suggestions improve as your labeled dataset grows - unleashing the, much sought after, virtuous cycle of the data flywheel effect in Vision AI. They are also plug-and-play, only requiring annotated data. The rest, we take care of in the background. For you, that means access cutting-edge Machine Learning technology to speed up your annotation process, without any hassle.
The core aspect of Hasty that brings you ‘Automated Labeling’ is a range of AI-powered assistants for every project. In essence, we build a neural network while you label that actively learns and progressively takes over the manual work for you.
We migrate through 3 levels of automation in annotations (check one of our older posts “Levels of annotation automation” for more details). You start with labeling a single object at a time with our blend of tools for level 1 automation.
After labelling ten images manually, you unlock our level 2 AI assistants, that help you label. Before they’re unlocked, the assistants learn on the first 10 manually labelled images and they retrain as you add more data - bringing active learning to your project. The results will not be perfect at the start, but the more data you add, the better the results. Going from 10 to 100, to 1000 annotated images you will see the degree of automation provided increasing, allowing you to improve annotation speed by 5-20x until you get to the point that with a single stroke of a button, you can annotate a whole image.
Once we see that level 2 is working well for you—by monitoring key metrics of the assistants and looking at user behavior patterns—we activate level 3 annotation. This allows you to batch process your entire dataset in one go with our assistants. Before running automated labeling, we give you an estimation of the credit cost based on the nature of the data to that point in the project. With this, you can go from labeling 100 images in a day at the start of a project to tens of thousands in an hour in the end, improving your efficiency hundredfold.
Many users come to Hasty after labeling many images with bounding boxes somewhere else but still aren’t getting the results they are looking for in their model. If you’re one of those, then this feature is for you. With Box to instance, you convert all the boxes in an image to pixel-wise masks or polygons with a single button click. As with our other assistants, the model is actively learning. Don’t worry if this conversion is not perfect at first, we have found that this assistant’s learning curve is extremely steep.
| Assistant | Task automated | Great for |
|---|---|---|
| Image tag prediction | the application of tags to an image to classify the nature of the entire image | capturing contextual information of the image scene |
| Attribute prediction | the assignment of attributes to a class or property to specify metadata about that class or property | adding another logic-layer to your model to increase accuracy |
| Label classification | our assistant picks the label class for you automatically | projects with more than 5 classes |
| Object detection assistant | automatically applies bounding boxes to target objects in an image | projects with many, clearly visible objects per image |
| Instance segmentation assistant | automatically applies pixel-wise masks or polygons (exact outlines) to target objects in an image | projects with many detailed objects and occlusions |
| Semantic segmentation assistant | automatically applies a semantic class to pixels that make up the relevant regions on an image | projects with defined regions of non-discrete classes (sky, road) |
| Box to instance | converts every bounding box in an image to instances (pixel-wise masks or polygons) | converting already bounding boxes to segmented masks |
Hasty has a set of APIs (https://docs.hasty.ai/reference) that allow you to automate repetitive tasks such as uploading images, downloading annotations, retrieving project metrics, etc.
We have many requests from our data-sensitive customers, looking for convenience to simply “hook” into their Amazon S3, Google, or Azure storage. With the launch of Pro plans, this is now easily possible. By submitting signed URLs we directly fetch the image from your cloud environment, saving you even more time.
Additionally, if you would like to mount your bucket directly to Hasty to keep the permanent storage in your environment, we can make that happen too.
As long as you are manually labeling data, Hasty remains free. We only charge credits when we automate work for you. Credits are our in-tool currency. Use your credits on certain actions in the tool, such as using one of our AI assistants to automatically label data, clean data, or rerun model training.
Everyone who signs up for Hasty receives up to 3000 free credits for the first 2 months and then 300 per month thereafter. Use your free credits to experiment and test our AI tooling and try out our features before you pay anything. Free credits also give you a chance to understand your credit consumption needs before deciding which plan works best for you.
Are you interested in estimating how many credits you would need? Try our pricing calculator.
We have also launched workspaces where billing and costs are linked to a workspace as an entity. This means that you can buy plans for a workspace and those credits are then consumable by all the projects within that workspace. So, if you have a budget you would like to keep control over, you can do so by setting up a workspace and put all the relevant projects in that one workspace.
To collect all of your projects in a workspace, you can move your free projects into a workspace for your team to work on. From that point, the credits consumed for those projects will come from the credits allocated to the workspace.
You can add an unlimited number of users to your workspace and control the permissions each user can have in that workspace, including assigning billing ownership, giving the ability to create new projects, add credits to that workspace, or purely viewing rights. Further, it is still possible to add users directly to a project within your workspace, without adding them to the workspace itself. This would result in that user being able to access only that project and not all of the projects in your workspace. We believe that with workspaces we offer the most flexible and powerful control of user access and permissions found among annotation tools today.
Finally, within the project settings, your team is able to generate reports and get an overview of key metrics in the project. You can break down performance metrics on a user, project, weekly or monthly level.
Only 13% of vision AI projects make it to production. With Hasty, we boost that number to 100%.
Our comprehensive vision AI platform is the only one you need to go from raw data to a production-ready model. We can help you with:
All the data and models you create always belong to you and can be exported and used outside of Hasty at any given time entirely for free.
You can try Hasty by signing up for free here. If you are looking for additional services like help with ML engineering, we also offer that. Check out our service offerings here to learn more about how we can help.
% of vision AI teams are not because data is their bottleneck. We fix this.
