Deterministic Analog Clock Reading Algorithm (Angle-measured, Floor-based) #2861
Description
Confirm this is a feature request for the Python library and not the underlying OpenAI API.
- This is a feature request for the Python library
Describe the feature or improvement you're requesting
Many failures in analog clock reading appear to come from heuristic or visual-approximation approaches.
I propose a deterministic computer-vision + geometry pipeline that eliminates guessing and rounding:
• Define 0° as vertical-up direction from clock center.
• Detect clock center and hand line segments.
• Measure exact angles of each hand using vector math.
• Convert angles to time using floor-based cumulative rules:
• minute = floor(theta_m / 6)
• hour = floor(theta_h / 30)
• No rounding or “closest number” heuristics.
• Optional consistency check for continuous hour hands:
theta_expected = 30h + 0.5m (+0.5/60*s)
This approach is deterministic, explainable, and robust across photos, icons, and illustrations.
It also supports jump-hour-hand clocks and clocks without explicit 12 o’clock markers.
Goal: eliminate the “ChatGPT cannot read clocks” failure class with a stable geometric solution.
Additional context
-south korea, cho hana