About the series
As you know, we at Humans in the Loop have a great love and appreciation of a well-designed annotation tool. After the great feedback on the reviews we published of our the best platforms on the market here and here, we decided that it’s time for a deep dive in some of our all-time favorites!
This article is the fourth from a series of 10 reviews which will be published each week. Our first three tool reviews on Supervise.ly, TrainingData.io and Annotate.Online can be found here, here and here. Soon we will be uploading the links to the other articles as they are released.
The whole series is based on the premise of transparency and honesty and none of these reviews are sponsored. They are just our way to give props to the best teams out there working on making annotation easier for AI teams, and to share some of the know-how that we have been accumulating over the past few years as a professional annotation company.
As in previous reviews, our parameters are:
- project management
If you have additional questions or want to get in touch with us to beta test or feature your tool in an upcoming article, feel free to email us at firstname.lastname@example.org!
Based in Berlin, Hasty.ai began in January 2019 at a hackathon organized by the industrial tech company builder Wattx. The core team has a great combo of skills in venture building, computer science, and UX/UI design and they are continuously rolling out new features. One of the main applications of the tool is in the field of AI for manufacturing given the background of the company but its features are incredibly useful for any other computer vision domain.
The entire annotation suite of the tool is available for free use without limitations and it can accommodate as many users as needed. For automation tools, the platform has an interesting credit score system (explained below), which is one of the most sensible pricing models we have come across in the domain.
Hasty knows that there is no “one size fits all” in the smart tools and that they depend on what type of image you are dealing with its resolution, its clarity, etc. so they offer a wide range of tools with customizable thresholds:
The platform offers exporting the annotations in formats such as COCO, JSON and PNG images. Moreover you can upload pre-annotated projects.
Hasty still doesn’t have all of the features necessary to enable labeling at scale with large teams but we are looking forward to some new features which will be rolled out soon, such as annotator workflows and queues. Another field where we would like to see improvements is the dataset management suite such as the option to remove or move images from one dataset to another, export only parts of a project, etc.
What comes in handy currently is the user roles and permissions panel which allows you to customize existing roles (owner, admin, supervisor, labeler, view only) or to add new ones. Another useful feature is the ‘Project summary’ where you can preview the number of images based on their status (in progress, to review, skipped, done) and the number of instances from each class for balancing purposes.
In addition, an option all of our supervisors love is the ‘Manual review’ panel where they can view each instance which was created in the project and very easily delete it or assign it to another class if it’s wrong, instead of going through the dataset image by image. Instances can be filtered by class, labeler, image status or by image name.
Automation is where Hasty.ai shines. While many platforms are focusing exclusively on smart segmentation features, Hasty has developed a series of AI Assistants which are powered by Active Learning and are trained as you annotate the data.
For bounding boxes, after labeling as little as 10 images and setting them as ‘to review’ or ‘done’, users can start applying the ‘Object Detection’ assistant. Throughout the project, the assistant is retrained multiple times based on the new annotations that the user creates, and especially when the data is homogenous, it can save an enormous amount of time. It can also be coupled with the ‘Class prediction’ feature which learns to automatically assign classes to the annotations in the project.
For polygons and semantic segmentation, the process is very similar and users can choose the ‘Instance’ and “Semantic Segmentation’ AI Assistants. After you have annotated ‘enough’ images (depending on the complexity of the dataset), you can apply the model that is trained in Hasty’s backed on the entire dataset – and in contrast to other platforms which only support generic models, in this case it’s an entirely tailor-made model based on your own annotations and classes.
These automation tools, in addition to data storage, are used on the basis of credits. Users get 3000 free credits per month to run and apply smart tools and automation in their projects, and they can also purchase additional credits from Hasty in the range of 0.01 EUR per credit.
Hope this was helpful! If you are working on an AI project and are currently reviewing which tool might be the most appropriate for it, get in touch with us and we would be happy to have a call and advise you on the best way to build your pipeline.