Here we will describe the few steps that will help you discover all the parts of Picsell.ia, this will follow the onboarding that you should see when you first connect to the platform.
Here you can add people to your Organization so you can later collaborate with them on common projects and give them a role so you are sure that they only have access to intended information, data...In this tutorial we will see :
What is a Project
What is an Experiment :
What is the Datalake
What is a Dataset
What is an Organization
Welcome to Picsell.ia
The first thing you see when landing on Picsell.ia is your homepage and it should look like this
To start our tutorial series, let's click on the 'Start' button in the upper card
1. What is a Project ?
In the first step of our tutorial, we are going to see what is a project, if you have followed the introduction, you should now see a page like this
This is what we call a project overview.
It is the page that summarizes all of the asset that belongs to or can be used in a project.
The two most important things to notice here are the following :
In Picsell.ia, your data (images + annotations) are completely independent of your projects. But to help you organize your work, we allow you to attach some datasets to your projects so you don't have to search within all your data when looking for something in particular inside a project.
The Datasets section in your project overview indicates the list of datasets attached to your project, allows you to attach a new dataset, and to set a dataset as default. Setting a dataset as default is telling your experiments (we will look into this later) to use it as a default source of data in case nothing is specified.
As you might have guessed, this is the list of the latest experiment created in this project.
From here, you can click on 'see all' to access the whole list, click on one row to dive into one experiment, or create a new experiment by clicking the green button in the upper-right corner.
The Dataset column tells you that a dataset is attached to an experiment and if you click on its name (in green here) you will be able to access the dataset directly.
Now let's click on the first row (my first experiment) to see what experiments are made of.
2. What is an Experiment ?
Experiments are entities grouped under a project. An experiment is the place where you can store files, log your results in interactive charts, dive into your train/test/eval sets of data (when linked to a dataset), store and display server logs, and even run trained model.
If you have followed the previous step, you should now be on a page that we call the Experiment Overview, that displays everything that you have logged.
This header summarize the core informations of your experiment such as the owner (the user that created the experiment), the dataset (if any), but also the most important metrics that you need to see in one glimpse (e.g. final accuracy, loss...) and any values you have logged (see this section for more information about logging to an experiment)
As you can see here, you have multiple options available if you want to save data and share it with your team on Picsell.ia.
For each file you have access to several information :
Its name (unique, used to retrieve the asset)
Its filename (for example main.py)
The update date
The upload data
The size (small if < 5Mo, large if not)
To download a file you can click on the icon on the upper-right corner of the desired card
As for the data, we have a namespace for files that allows us to automatically recognize some of them to perform actions (such as deploying a model, launch a training...). Please refer to this page to see the reserved namespace.
You may have notices that we didn't covered all the tabs of the header, such as Telemetry or Test, in this tutorial. But don't worry, we wanted to keep it simple for now but we will see what's in it later.
3. What is the Datalake ?
No, you are not dreaming, at Picsell.ia we make available to you your own private Datalake 😎
We know you can't wait to discover it so here is what it looks like :
The Datalake is the place where you can upload, import, tag and sort all of your data.
This table lists literally ALL your assets. But to help you dive into your data and find what you need, we have added a powerful search input where you can search for tags and filenames.
If you select some assets by ticking some checkboxes in the table you will see something like this.
This is the list of operation that you can perform on your data, here you can :
add assets (upload or import images)
add tags (to the selected images)
create new datasets (so you can select and version data easier)
import data (in an existing dataset)
create a new version (of a dataset, with new assets)
To see the details of how each operation works please refer to the corresponding page of the documentation.
Now that we have seen how data is stored globally on Picsell.ia thanks to the Datalake, let's now dive deep into datasets, what you can with them and why it's awesome.
4. What is a Dataset ?
If you are familiar with AI and Data Science, you have inevitably heard of a 'dataset' some day in your life.
But as you may not know how Picsell.ia works, we will still explain what a Dataset is here.
This is a Dataset version page. At first, it looks quite similar to the Datalake page, let's see what we really have here.
The first use of a dataset is the ability to annotate your data.
In this screenshot, you can see the same kind of table as in the Datalake, except that there are new columns.
Indeed, by clicking on the green button at the end of each row, you can access our labeling interface that allows you to create ground truth for your future experiments (we will see the interface later).
As we can see, some images are annotated (the Instances column indicates the classes labeled in the image).
You can access your Dataset analytics page by clicking on the analytics tab
On this page you'll be able to visualize your :
Dataset repartition ( number of objects by class )
Object by Image distribution
Object width distribution
Object height distribution
You can access your Dataset settings by clicking on the settings tab
Thanks to your settings page you'll be able to:
Configure/modify your labels
Import annotations from other dataset version
Configure/modify questions and answers asked during the annotation process
5. What is an Organization
Yes, you guessed it, this is the last part of this tutorial about Picsell.ia (but please don't cry we have a lot more to show you in the other section of the docs ! 😇)
When you create an account on Picsell.ia (we are glad you did) we automatically create your own Organization.
If you take a look at the main sidebar of the platform, at the top you should see a collapsible link with the name of your organization (which is at first the same as your username).
Here we have access to several links that will redirect you either to your profile settings or to your Organization settings
Let's click on the Informations link to see what Organization are here for.
These tabs are your Organization's settings. From here you can
Add members to your organization
Have access to your billing details and invoices
Update your subscription
For now, we will just learn how to add members to our Organization
If the person you are looking for already has an account on Picsell.ia, you can enter his/her email or username and it will send them an invite to join your Organization.
If not, you have to enter a valid email address and we will send them an email containing an invitation to join Picsell.ia and your Organization.