" %}
Example Loading Image Classification Interface
{% endembed %}
## Installation
```bash
npm install universal-data-tool
```
## Usage
The `` component has every interface baked in. Just provide a dataset in the `dataset` prop to view any dataset!
```javascript
import React from "react";
import UniversalSampleEditor from "universal-data-tool";
export default function App() {
return (
{
// Called when user hits "Save", "Next", or "Prev"
}}
// when data is saved this is called
onModifySample={(sampleIndex, newSampleData) => {
// do something
}}
/>
);
}
```
# Getting Started with UDT Courses
Create Training Courses for Labeling Datasets
{% embed url="" %}
This video has everything you need to know!
{% endembed %}
## Creating a Course
{% hint style="warning" %}
You'll want to have a dataset before you begin! No labels are needed.
{% endhint %}
You can create UDT courses online in a couple easy steps. First navigate to [universaldatatool.com/courses](https://universaldatatool.com/courses), then click "Create a Course".
Now you just need to upload your dataset. You can drag and drop or select your dataset here. The samples will be used to construct tests!
You should now be inside the course editor!
## Editing a Course
Every course has an Instructor URL and a Student URL. Your Instructor URL has a secret code in it that allows you to come back, view passing students and edit the course. So don't lose it! After you create a course it should be in your address bar.
{% hint style="info" %}
Here's an example of an Instructor URL:
And here's an example of a Student URL:
See how the Instructor URL is longer and has an "edit\_key"?
{% endhint %}
We can now add sections, tests, exercises, quizzes and explanations to create a full training course.
### Creating a Course Item
Anytime you see this icon, you can create a new course item.
### Markdown Descriptions
It's just markdown! Put whatever you want inside to explain the motivation of your dataset or specific details. Maybe start with a general overview, then do a deep dive in a second section.
### Question / Quizzes
Quizzes are used to make sure that your labelers read the preceding paragraph.
### Tests / Exercises
**Tests and exercises are the bread and butter of a UDT course.** They make sure your labelers actually know how to label with real examples.
#### 1. Select the samples
#### 2. Edit the Solution
Here you can add solutions to the dataset. **If your dataset already has labels, they'll automatically be imported.** Your labelers will be tested on these solutions!
#### 3. Configure the Test
Configure your test with instructions, make it a practice exercise etc.
## Sharing a Course with Labelers
Share your course using the "Go to Course" button at the top of the page. That will take you to a page that you can send to all of your labelers.
After a student successfully completes the course, they'll be prompted to enter their email. Their email will now appear in the "Passing Students" dialog.
## Running On-Premise
Check out the [github repository for on-premise instructions](https://github.com/UniversalDataTool/courseware)!
# Frequently Asked Questions
FAQ for the Universal Data Tool
## Why are fractional coordinates used instead of pixel coordinates in the Universal Data Tool?
Pixel coordinates can often be helpful for drawing or extraction part of an image. However, fractional coordinates, meaning coordinates in the range `[0,1]` are used for representing the location on an image in the UDT for the following reasons:
* It is resilient to image resizing/changes to size in the underlying image source
* Most images are rescaled prior to being put into a machine learning model, at which point a pixel coordinate transformation would be necessary and cumbersome
* It better supports "massive" image scenarios, such as maps or microscopic slides, that have variable pixel accuracy
If you need to convert UDT fractional coordinates into pixel coordinates, you can do it quickly with[ udt-to-image-position](https://github.com/UniversalDataTool/udt-to-image-position).