The Plainsight Way: Developing Computer Vision Applications

Introduction

The "Plainsight Way" is a structured and scalable approach to developing computer vision applications. It is designed to help Embedded Vision Providers (EVPs) build, deploy, and manage sophisticated vision pipelines efficiently.

Plainsight provides a robust ecosystem of tools, templates, and support services, enabling EVPs to focus on their core expertise—integrating and customizing solutions to meet end-customer needs.

Responsibilities Overview

The development of computer vision applications requires a collaborative effort between Plainsight and the EVP. Below is a high-level breakdown of each party's responsibilities.

Plainsight Responsibilities

Plainsight provides the foundational infrastructure, tools, and best practices necessary for building, testing, and deploying vision applications efficiently.

• Utility Filters: A library of pre-built filters designed to perform common transformations (cropping, resizing, filtering, format conversions, etc.).
• Custom Filter Development Support: A framework for building and publishing custom filters, allowing EVPs to extend functionality.
• Computer Vision Recipes: Pre-configured workflows that handle routine vision tasks, such as image collection, pre-processing, and model execution.
• Protege – Model Training and Evaluation: A platform to train, evaluate, and benchmark custom models using labeled datasets.
• Testing and Deployment Tools:
  • Quality Gate Filter Output Verification Tool for performance benchmarking and quality assurance.
  • Jester Data Ingestion Service for automated data handling and pre-processing.
• Templates and Examples: Example Docker Compose configurations and deployment templates for rapid prototyping.
• Cloud-based Orchestration and Deployment: Services and tools for running Filters in cloud environments via Vision Flow, Vision Stream, and Vision Edge.
• Ongoing Support and Consultation: Assistance with debugging, performance optimization, and customization.
• Future Enhancements: Features like data lineage tracking, dataset curation, and advanced filter orchestration will be integrated into the platform.

EVP Responsibilities

As the primary integrator, the EVP plays a crucial role in adapting Plainsight’s technology to real-world use cases and customer environments.

• Providing Data: Ensure the availability of quality data for model training and evaluation.
• Filter Integration and Customization:
  • Configure and manage Docker Compose files to deploy and orchestrate Filters.
  • Customize utility filters for domain-specific applications.
• Processing Subject Data: Handle the results generated by application filters, integrating them into customer applications and analytics pipelines.
• Building Custom Filters: Develop specialized filters using Plainsight’s framework to extend platform capabilities.

Plainsight Tools and Their Roles

Tool	Purpose
Jester	Automates data ingestion, preprocessing, and management.
Protege	Model training and evaluation platform.
Quality Gate	Benchmarking and verification tool for filter outputs.
Filter SDK	Framework for building and deploying filters.
Vision Flow, Vision Stream, Vision Edge	Orchestrate filter execution in cloud, batch, and edge environments.
Telemetry & Lineage Tools	Monitor filter execution and track data lineage.

By following the Plainsight Way, EVPs can efficiently develop, deploy, and optimize computer vision applications, ensuring high performance and reliability in real-world deployments.