Hasty is an image annotation tool for creating ground-truth datasets to be used in machine learning applications. Unlike many similar tools, Hasty uses machine learning in the tool itself. This allows you to annotate exponentially quicker than before.
To get started with Hasty, go to https://app.hasty.ai/ and sign up either by using your email or by logging in through your Google account.
If you want to know more before signing up, you can go to https://hasty.ai .
To log in or sign up with your Google account, just click the "Login with Google" button on the first screen. After clicking the button, a new window will open where you can fill in your Google account and password. Do so and you will be logged in (if you have used Hasty before) or signed up (if you haven't used Hasty before).
To signup/login with an email address click the "Login with email" button. On the next screen, you will be asked for an email, fill in the email address you want to use and then click on "Continue". When you do so, an email containing a link will be sent to the email you specified. Open up the email, click the link, and then you will be logged in (if you have used Hasty before) or signed up (if you haven't used Hasty before).
After you have logged in, you will find yourself in the project overview screen:
From here, you can either create a new project, change existing projects, or start testing out our already existing demo project.
Do you know that we have a guide for getting you started with Hasty? As the creative masterminds we are, we named it "Getting started with Hasty". Check it out to learn more about how to use Hasty.
We already mentioned that Hasty uses machine learning to speed up the process of annotation. However, our AI tools are not available when you are starting a new project. For these tools to work, we first need you to create enough annotations manually so that we can generate a model.
To activate the tools, you will need to completely annotate either ten images (and set their status to "done" or "to review"), or in the case of the Class Predictor Assistant, create 25 annotations.
After giving the model(s) some time to train, you will be able to start using our AI tools. Of course, suggestions might not be perfect from the beginning but the models will improve as you annotate more images. In our experience, annotating around 10 annotations per label class is generally a good rule-of-thumb for achieving useful results.
Can’t find what you are looking for in our documentation? Get in contact with me, Herbert Janssen, community manager here at Hasty by emailing [email protected] if you have any questions or need help.
You can edit an existing label class by hovering over it with your mouse pointer and then click. This will open up an edit overlay where you can change the name, color, and type of the class.
You can also change the type of a label class by dragging it from one type to the other like so:
There are two different types of label classes. Object and Semantic.
Object classes can be used in all types of annotation. You use object classes to annotate "things" or "objects" - meaning any actual object or animal with a discernible and coherent shape.
Examples of object classes are: Melons, bicycles, houses, and hedgehogs.
Semantic classes are mainly used when doing semantic segmentation and panoptic segmentation and are used to mark "stuff" - meaning anything that is not an object or an animal.
Examples of semantic classes are: Sky, grass, and water.
When exporting using our semantic segmentation format, we automatically add a background class to all pixels that haven't been annotated by users - so you don't need to create a background class yourself
On this screen, you can create and remove datasets in a project. Datasets are collections of images, not unlike folders in your operating system, where you can collect images.
The usage of arranging your images in specific datasets can be many, but to give you some examples:
- You can create individual datasets for images taken in different settings. This will allow you to concentrate on a specific type of data.
- Example: Stockholm, Berlin, Hong Kong, Buenos Aires
- You can create individual datasets for specific annotators or team of annotators. This will allow you to assign a dataset to a person/team.
- Example: Team Blue, Team Pink, Team Rocket
- You can create a new dataset on every new import of data. This makes it easier to see when a specific section of your project data has been completed while also improving quality control workflows
- Example: 19th of September 2019, 5th of October 2019, 2nd of November 2019
You can create a new dataset easily by going to the form field at the top of a page, then write the name you want your dataset to have, and then press "enter".
You can remove a dataset by hovering with your mouse over a dataset, and then click the cross that appears in the top right corner.
Before you upload, click on the dropdown menu above the drop area to select to which dataset you want to upload your images. Please note that you can only upload images to one dataset at a time.
You can upload images to a dataset by either dragging files or folders into the drop area or by clicking the link underneath the cloud icon and then select using your file explorer which files/folders you want to add.
Currently, we can only support a limited amount of file types. If your file is not in one of these formats you will not be able to add it to Hasty. If your file type is supported but you still get error messages, contact us directly and we will solve your problem as quickly as we can.
When you add a new user, they will receive an email that they have been added to the project.
You can also remove a user by hovering over the name and then pressing the x that appears in the top right corner.
Finally, you can edit the role of a user by clicking on the name and then changing the assigned role in the modal dropdown menu.
User roles are a function put in place so that you can easily set up permissions for many users at once. It allows you to specify what a specific role can and cannot do. When you have it set up as you want, you can then easily assign any user that specific role. There are 5 default user roles that you can use from the start. Those are:
- Project owner: The owner of the project. This will always be the creator of the project. The project owner is the only one that can change permissions for users.
- Administrator: Your typical administrator account. In our default setup, the admin can do anything except for editing user roles.
- Supervisor: In larger projects, it is common to have specific supervisors that check the quality of annotations regularly. Users with this role can't administrate the project but can see the project summary report and set an image to the image status "done" unlike a...
- Labeller: A labeller can only view and annotate a project.
- View only: Mainly a role for clients. Allows view-permissions to a project but doesn't let you change anything.
The intended workflow with this setup is that a labeller annotates an image and then set the status of the image "to review". A supervisor checks the images and, if satisfied, updates the status to "done". Administrators and the project owner can participate in both steps, and can also change the project settings as they go.
However, we know that workflows changes from organisation to organisation. That's why user roles are completely customisable allowing you to set up your project to fit your workflow.
To add a new role, click the "Add new" button. Then pick a color and a name for the role.
User role colors are used throughout the tool as a way of allowing you to see what role a user have easily. When creating a new role, try to pick a color that is not too close to the color of an already existing role.
You can edit a role by clicking on the name at the top of the permissions table. This will open up a modal in which you can change the name and the color of the role.
You can remove a role by hovering over its name, and then clicking the x that will appear in the top right corner.
Below the role names, you will have a series of boxes. These are permission boxes. By clicking one of them, you can add/remove a permission. You add a permission by clicking an empty box and remove a permission by clicking on a box with a checkmark in it.
When editing permissions, don't forget to save your changes by clicking the "Save changes" button at the bottom of the page.
The permissions that are available in the tool are organised in four different categories. The categories and permissions are:
- Administrate: Permissions tied to the administration of a project.
- Edit project classes: Permits creation, editing, and removal of label classes
- Upload images: Permits users to upload images to a project
- Export: Permits users to export annotations from Hasty
- Invite/Edit/Remove users: Permits user to invite, remove, or edit the roles of users
- Import: Permits users to upload own annotations to the project
- Manage roles: Permits users to manage roles - including removing/adding new roles and changing the permissions of existing roles. For now, this is fixed so that only project owners can manage roles.
- Annotation: Permissions that impacts the annotation environment
- Annotation: Roles with this permission can add, delete, and edit annotations in the annotation environment.
- Supervision: A permission that allows a user to set an image to "done"
- Project summary: Roles with this permission can see the project summary report.
Herbert Janssen, community manager here at Hasty by emailing [email protected] if you have any questions or need help.
Updated 5 months ago