Synthetic Visual Dataset Prompting Framework and Curation Report

Role: Lead Vision-Language Data Strategist and Prompt Systems Architect

Context

• Project Title: {{project_title}}
• Target Vision Task: {{target_vision_task}}
• Base Generator Model: {{base_generator_model}}
• Diversity Dimensions: {{diversity_dimensions}}
• Camera & Lighting Specifications: {{lighting_and_camera_specs}}
• Curation & Quality Filters: {{curation_filtering_criteria}}

Task

Author a comprehensive synthetic data generation specification report for {{project_title}} that details programmatic prompt syntax, parameter variability ranges, and post-generation curation rules using {{base_generator_model}} to support {{target_vision_task}}.

Method

1. Analyze {{target_vision_task}} requirements to identify edge cases, bounding box balance, and class representation targets.
2. Establish combinatorial prompt generation templates that cycle through {{diversity_dimensions}} systematically.
3. Embed rigorous optical presets derived from {{lighting_and_camera_specs}} into prompt suffix modifiers.
4. Design stratified prompt matrices that control background density, occlusions, and object orientation.
5. Formulate automated negative prompt configurations to systematically suppress non-target distribution artifacts.
6. Detail programmatic post-generation filtering rules based on {{curation_filtering_criteria}} to eliminate degenerate samples.
7. Define human-in-the-loop review protocols for boundary case validation and statistical distribution tracking.
8. Construct a step-by-step pipeline blueprint linking prompt generation, image synthesis, and automated annotation.

Constraints

• MUST define exact parametric placeholders (e.g., {environment}, {subject_pose}, {lighting_angle}) within prompt templates.
• MUST NOT permit unquantified diversity attributes; all variations must have explicit token sets.
• MUST establish programmatic image acceptance and rejection thresholds aligned with {{curation_filtering_criteria}}.
• Prompt modifiers MUST maintain photorealistic optical coherence without introducing digital painting artifacts.

Output format

Format the output as a technical dataset engineering report structured with these exact sections:

1. Pipeline Overview & Objective: {{project_title}}

2. Combinatorial Prompt Engineering Architecture

3. Optical & Lighting Prompt Token Dictionary

4. Diversity Matrix & Permutation Rules

5. Automated Curation & Artifact Elimination Protocol

6. Quality Assurance & Pipeline Validation Metrics

Expected length: 1,300 to 1,900 words.

Self-review

1. Ensure all variables from {{diversity_dimensions}} are mapped to concrete, executable prompt tokens.
2. Confirm that every prompt template contains strict optical parameters from {{lighting_and_camera_specs}}.
3. Validate that curation thresholds explicitly address requirements for {{target_vision_task}}.