Skip to main content

Best practices

Proper project organization is essential for efficient dataset management and model development. Follow these guidelines to structure your work effectively.

Project structure and organization​

Define project scope early​

Before creating a project, document:

  1. What are you detecting? (objects, defects, conditions)
  2. What type of task? (classification, detection, segmentation)
  3. Success criteria (accuracy targets, performance requirements)
  4. Production environment (lighting, camera setup, speed requirements)
Avoid Scope Creep

Resist the temptation to combine multiple unrelated tasks into one project. This can complicate training and reduce model performance.

Use clear, descriptive names​

Project names should be informative and follow a consistent naming convention.

Good naming practices:

  • ✅ SilkRoad - Bottle Orientation Detection v2
  • ✅ SilkRoad - Product Segmentation 2024
  • ✅ SilkRoad - Defect Classification ProductionA

Avoid:

  • ❌ project1, test, new_project
  • ❌ Vague names that don't describe the task
  • ❌ Special characters
Naming Convention

Use capitalized words separated by spaces, include client name, object, task, and variant when relevant: [Client] - [Object] [Task] [Version/Variant]

Organize by use case or product line​

Create separate projects for different detection tasks or product lines, even if they share similar objects.

Example structure:

PROJECTS/
├── bottle orientation detection/
├── bottle defect inspection/
├── cap presence verification/
└── label quality control/

This separation allows independent model training and deployment cycles.

Collaboration and team management​

Assign clear responsibilities​

For team projects, define roles clearly:

  • Project owner - Overall responsibility and decision-making
  • Annotators - Perform labeling according to guidelines
  • Reviewers - Quality control and validation
  • Model trainers - Configure and train models
Role Definitions

Follow the Least privilege principle, grant only the access level necessary for each of the member's responsibilities.

Regularly review access and permissions​

Ensure team members have appropriate access levels based on their roles. Periodically audit permissions to maintain security and efficiency.

Project management checklist​

Before starting:

  • Clear project name and description
  • Defined objectives and success criteria
  • Complete label schema planned
  • Labeling policy documented
  • Team roles assigned (if applicable)