Generating Interfaces for Custom Algorithm

Bilayers simplifies the process of creating web-based (Gradio) or Jupyter Notebook interfaces for any deep-learning cell segmentation algorithm. Instead of writing extensive UI code, you only need to fill out a structured YAML file—just like filling out a Google Form!

Before You Begin:¶

To successfully generate an interface, familiarize yourself with the key Bilayers components:

1. Supported Interfaces
   Bilayers currently supports the following interface types:

   • Gradio - A no code web UI

   • Jupyter Notebook - Interactive notebooks

2. Understanding config.yaml requirements The config.yaml file defines the behavior of your interface, including input/output handling and algorithm execution. Each section of this file serves a distinct purpose:

   • citations - Reference publications or software licenses

   • docker_image - Specify the pre-built Docker image

   • algorithm_folder_name - Define a workspace directory

   • exec_function - Configure how the algorithm runs

   • inputs - Define expected input files/data

   • outputs - Specify expected results

   • parameters - Configure user-adjustable options

   • display_only - Read-only UI fields

3. Choosing the right base docker image Your algorithm’s Docker image must meet Bilayers’ compatibility requirements. Follow these guidelines to ensure seamless integration:

   • Requirements for the Algorithm (Base) Docker Image - Must include Python, package managers, and other essentials

   • Adapting Non-Compliant Base Images - Steps to modify an existing image

4. Steps to create your custom Algorithm’s Interfaces Once you’ve selected an interface type, configured config.yaml, and prepared a compatible Docker image, follow the Step-by-Step Guide to generate your UI automatically.