DeepEM Playground

Getting Started

Our DEEP-EM playground leverages two powerful external tools to ensure accessibility and ease of use for everyone:

Lightning AI: This platform provides seamless access to our use cases, enabling effortless collaboration and utilization of GPU environments directly within your browser.
CVAT (Computer Vision Annotation Tool): This tool simplifies the process of annotating your own data, making it straightforward and efficient.

In the sections below, we provide a brief introduction to help you get started with each of these tools.

Before starting off with your first use case, make sure to read the content at Workflow carefully.

1. Register with Lightning AI

In order to use the provided playground within Lighting AI Studios, a registration is required. Please setup an user account by following this link. Registration may take up to 2-3 days.

Pricing

Lighting AI Studios offers three plans: one free plan and two paid plans. For details on the plans please follow this link

By completing the registration steps outlined above, you'll gain access to Lightning AI Studio's free plan, which includes 15 compute credits per month. This plan allows the use of a free, CPU-based Studio for up to 4 hours at a time, with the option to restart it after each session to continue using it at no cost. For deep learning tasks, we recommend GPU-based Studios, which require compute credits. The number of credits consumed depends on the type of GPU selected. All use cases in our playground are pre-configured with a default GPU setup optimized for computation, ensuring smooth and efficient performance. Additional credits can be added as needed using the pay-as-you-go model.

2. Use Case

In the following, we provide a step-by-step guide on how to get started with a use case. This guide will walk you through the process of exploring a specific use case, showing how to apply it to your own data and research needs.

3. Upload Data

Our standardized workflow enables the adaptation of Use Cases within their Primary Focus without the need for coding. Instead, you only need to replace the data being used. To do so, simply upload your data to Lighting AI. A step-by-step guide will follow to walk you through the process. For more details, please see here.

Notes on Data Ownership & Security with Lightning AI

Key Points

When using Lightning AI’s services, here are the important aspects related to your data:

Data Ownership: You retain ownership of your data and models (referred to as "Customer Models"). Lightning AI does not own them.
Access to Your Models: Lightning AI has administrative access to your models only for operational purposes (like support or troubleshooting), not for unauthorized use.
Performance Data: Lightning AI can use performance data (e.g., how well your models perform) for any purpose, but not your original models.
Security Measures: Lightning AI commits to using commercial-grade security measures to protect your data. If there's a breach, they will notify you promptly.

What Does This Mean for Your Sensitive Data?

While Lightning AI ensures they won't access your models for unauthorized purposes, they will have access to them for providing the services you requested, such as technical support. They are required to protect your data with appropriate security measures and notify you if there’s a breach.

Important: If you upload sensitive data, it’s recommended to understand their security protocols. You are also responsible for ensuring you have the legal rights to the data you upload.

Ultimately, if you are concerned about privacy, review their security policies and ensure your data handling complies with relevant regulations. You can find further details here and here.

4. Data Annotation

Most Use Cases require an annotated dataset to work. To be able to provide your annotated data in the correct format, we recomment using the CVAT (Computer Vision Annotation Tool). It is a free, browser-based tool for annotating images and videos, commonly used in computer vision tasks. It supports various annotation types like bounding boxes, segmentation masks, and points, and allows multiple users to collaborate on the same dataset for faster annotation. CVAT is especially useful for researchers working with electron microscopy (EM) data, as it is one of the few free tools that supports the .tif format. It also offers paid options for faster, AI-assisted annotation. Whether used directly in the browser or set up locally, CVAT is an easy-to-use, flexible solution for efficient annotation. In the following we will explain how to setup CVAT in the browser. If you require a local setup please see their documentation and follow: this link

For users seeking a more technical introduction, a comprehensive guide is available here: CVAT Getting Started.

Pricing

CVAT offers both cloud-based and self-hosted versions to suit different user needs. The cloud version, available via CVAT Cloud Pricing, provides a scalable, hosted solution, while the self-hosted version, detailed in CVAT On-Prem Pricing, allows for local installation and management. The free version includes the ability to manage up to 3 parallel projects and supports 10 tasks in total, with a maximum of 5 tasks per project. Tasks represent individual units of work within a project, and they can be split into multiple jobs for parallel data annotation by different workers, making it easier to handle large datasets efficiently.

Note on Data Privacy

When using the CVAT annotation tool in your browser, your data is uploaded to CVAT.ai's servers. According to their Privacy Policy, CVAT.ai is committed to protecting your personal information and outlines the rights you have under various data protection regulations. They state that your data is stored securely on their AWS instances and is not distributed further. If you delete a project or task, all related data will be deleted as well.

For more detailed information, we refer to CVAT.ai's Terms of Use and Privacy Policy. If you have specific concerns or require further clarification, it's advisable to contact CVAT.ai directly through their page.

5. Collaboration

At some point within the process you probably wish to share the work with fellow collegues. This could be for working together on improving the model training, or by allowing your collegues access to a model which you've trained. This can be achieved easily using teamspaces of the Lightning AI studio. For an introduction into what a teamspace is, please see here. For further details on how to enable sharing see below.

BibTeX

BibTex Code Here