Mobile UI — Screen Extraction and Indexing

Prompt

You are evaluating a proposed screen-extraction workflow for a fictional mobile application. The project contains approximately 40 flattened PNG comparison sheets, potentially representing more than 200 individual screens across phone portrait, tablet portrait and tablet landscape layouts. Your task is to demonstrate how you would execute a screen inventory and extraction workflow. Requirements: - Assign stable source IDs and screen IDs. - Identify individual screens inside multi-screen PNGs. - Record bounding rectangles in original image coordinates. - Preserve links to original source files. - Identify device type, orientation and layout. - Flag ambiguous boundaries and partial screens. - Detect apparent duplicates without silently removing them. - Produce machine-readable JSON. - Specify independent validation checks. Do not invent image observations, bounding rectangles or screen counts. If no actual PNG is accessible, explicitly distinguish what can be implemented from what requires image inspection. Deliver: 1. A proposed JSON schema. 2. The screen-detection and cropping algorithm. 3. A Python implementation approach. 4. Automated validation rules. 5. A list of visual ambiguities requiring human or AI judgement. Optimise for accuracy, reproducibility, low token consumption and processing hundreds of screens in manageable batches. Avoid unnecessary redesign, speculative metadata and additional infrastructure.